Date: Wed, 10 Aug 2022 15:38:17 +0200 Subject: Change dev branch used for docs, update static docs --- docs/dev/articles/FOCUS_L.html | 1239 ++++++++++---------- .../figure-html/unnamed-chunk-10-1.png | Bin 42203 -> 42478 bytes .../figure-html/unnamed-chunk-12-1.png | Bin 82006 -> 83170 bytes .../figure-html/unnamed-chunk-13-1.png | Bin 32751 -> 33165 bytes .../figure-html/unnamed-chunk-15-1.png | Bin 57939 -> 58691 bytes .../figure-html/unnamed-chunk-4-1.png | Bin 35238 -> 35677 bytes .../figure-html/unnamed-chunk-5-1.png | Bin 21999 -> 22137 bytes .../figure-html/unnamed-chunk-6-1.png | Bin 35640 -> 36101 bytes .../figure-html/unnamed-chunk-8-1.png | Bin 41399 -> 41430 bytes .../figure-html/unnamed-chunk-9-1.png | Bin 40983 -> 40780 bytes docs/dev/articles/index.html | 9 +- docs/dev/articles/web_only/benchmarks.html | 555 +++++++-- docs/dev/articles/web_only/dimethenamid_2018.html | 452 +++---- .../figure-html/f_parent_mkin_dfop_const-1.png | Bin 58963 -> 58608 bytes .../f_parent_mkin_dfop_const_test-1.png | Bin 59123 -> 58777 bytes .../figure-html/f_parent_mkin_dfop_tc_test-1.png | Bin 60654 -> 60062 bytes .../figure-html/f_parent_mkin_sfo_const-1.png | Bin 57222 -> 56694 bytes .../figure-html/plot_parent_nlme-1.png | Bin 59533 -> 59194 bytes 18 files changed, 1298 insertions(+), 957 deletions(-) (limited to 'docs/dev/articles') diff --git a/docs/dev/articles/FOCUS_L.html b/docs/dev/articles/FOCUS_L.html index 547ec630..610d1bdd 100644 --- a/docs/dev/articles/FOCUS_L.html +++ b/docs/dev/articles/FOCUS_L.html @@ -20,6 +20,8 @@ + +

-Laboratory Data L1

Laboratory Data L1 +

The following code defines example dataset L1 from the FOCUS kinetics report, p. 284:

-library("mkin", quietly = TRUE)
-FOCUS_2006_L1 = data.frame(
-  t = rep(c(0, 1, 2, 3, 5, 7, 14, 21, 30), each = 2),
-  parent = c(88.3, 91.4, 85.6, 84.5, 78.9, 77.6,
-             72.0, 71.9, 50.3, 59.4, 47.0, 45.1,
-             27.7, 27.3, 10.0, 10.4, 2.9, 4.0))
-FOCUS_2006_L1_mkin <- mkin_wide_to_long(FOCUS_2006_L1)

library("mkin", quietly = TRUE)
+FOCUS_2006_L1 = data.frame(
+  t = rep(c(0, 1, 2, 3, 5, 7, 14, 21, 30), each = 2),
+  parent = c(88.3, 91.4, 85.6, 84.5, 78.9, 77.6,
+             72.0, 71.9, 50.3, 59.4, 47.0, 45.1,
+             27.7, 27.3, 10.0, 10.4, 2.9, 4.0))
+FOCUS_2006_L1_mkin <- mkin_wide_to_long(FOCUS_2006_L1)

Here we use the assumptions of simple first order (SFO), the case of declining rate constant over time (FOMC) and the case of two different phases of the kinetics (DFOP). For a more detailed discussion of the models, please see the FOCUS kinetics report.

Since mkin version 0.9-32 (July 2014), we can use shorthand notation like "SFO" for parent only degradation models. The following two lines fit the model and produce the summary report of the model fit. This covers the numerical analysis given in the FOCUS report.

-m.L1.SFO <- mkinfit("SFO", FOCUS_2006_L1_mkin, quiet = TRUE)
-summary(m.L1.SFO)

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:39 2021 
-## Date of summary: Mon Feb 15 17:13:39 2021 
-## 
-## Equations:
-## d_parent/dt = - k_parent * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 133 model solutions performed in 0.032 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0 89.85  state
-## k_parent  0.10 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##                  value lower upper
-## parent_0     89.850000  -Inf   Inf
-## log_k_parent -2.302585  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC     BIC    logLik
-##   93.88778 96.5589 -43.94389
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##              Estimate Std. Error  Lower  Upper
-## parent_0       92.470    1.28200 89.740 95.200
-## log_k_parent   -2.347    0.03763 -2.428 -2.267
-## sigma           2.780    0.46330  1.792  3.767
-## 
-## Parameter correlation:
-##                parent_0 log_k_parent      sigma
-## parent_0      1.000e+00    6.186e-01 -1.516e-09
-## log_k_parent  6.186e-01    1.000e+00 -3.124e-09
-## sigma        -1.516e-09   -3.124e-09  1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##          Estimate t value    Pr(>t)    Lower   Upper
-## parent_0 92.47000   72.13 8.824e-21 89.74000 95.2000
-## k_parent  0.09561   26.57 2.487e-14  0.08824  0.1036
-## sigma     2.78000    6.00 1.216e-05  1.79200  3.7670
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   3.424       2  7
-## parent     3.424       2  7
-## 
-## Estimated disappearance times:
-##         DT50  DT90
-## parent 7.249 24.08
-## 
-## Data:
-##  time variable observed predicted residual
-##     0   parent     88.3    92.471  -4.1710
-##     0   parent     91.4    92.471  -1.0710
-##     1   parent     85.6    84.039   1.5610
-##     1   parent     84.5    84.039   0.4610
-##     2   parent     78.9    76.376   2.5241
-##     2   parent     77.6    76.376   1.2241
-##     3   parent     72.0    69.412   2.5884
-##     3   parent     71.9    69.412   2.4884
-##     5   parent     50.3    57.330  -7.0301
-##     5   parent     59.4    57.330   2.0699
-##     7   parent     47.0    47.352  -0.3515
-##     7   parent     45.1    47.352  -2.2515
-##    14   parent     27.7    24.247   3.4528
-##    14   parent     27.3    24.247   3.0528
-##    21   parent     10.0    12.416  -2.4163
-##    21   parent     10.4    12.416  -2.0163
-##    30   parent      2.9     5.251  -2.3513
-##    30   parent      4.0     5.251  -1.2513

m.L1.SFO <- mkinfit("SFO", FOCUS_2006_L1_mkin, quiet = TRUE)
+summary(m.L1.SFO)

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:18 2022 
+## Date of summary: Wed Aug 10 15:28:18 2022 
+## 
+## Equations:
+## d_parent/dt = - k_parent * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 133 model solutions performed in 0.031 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0 89.85  state
+## k_parent  0.10 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##                  value lower upper
+## parent_0     89.850000  -Inf   Inf
+## log_k_parent -2.302585  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC     BIC    logLik
+##   93.88778 96.5589 -43.94389
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##              Estimate Std. Error  Lower  Upper
+## parent_0       92.470    1.28200 89.740 95.200
+## log_k_parent   -2.347    0.03763 -2.428 -2.267
+## sigma           2.780    0.46330  1.792  3.767
+## 
+## Parameter correlation:
+##                parent_0 log_k_parent      sigma
+## parent_0      1.000e+00    6.186e-01 -1.516e-09
+## log_k_parent  6.186e-01    1.000e+00 -3.124e-09
+## sigma        -1.516e-09   -3.124e-09  1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##          Estimate t value    Pr(>t)    Lower   Upper
+## parent_0 92.47000   72.13 8.824e-21 89.74000 95.2000
+## k_parent  0.09561   26.57 2.487e-14  0.08824  0.1036
+## sigma     2.78000    6.00 1.216e-05  1.79200  3.7670
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   3.424       2  7
+## parent     3.424       2  7
+## 
+## Estimated disappearance times:
+##         DT50  DT90
+## parent 7.249 24.08
+## 
+## Data:
+##  time variable observed predicted residual
+##     0   parent     88.3    92.471  -4.1710
+##     0   parent     91.4    92.471  -1.0710
+##     1   parent     85.6    84.039   1.5610
+##     1   parent     84.5    84.039   0.4610
+##     2   parent     78.9    76.376   2.5241
+##     2   parent     77.6    76.376   1.2241
+##     3   parent     72.0    69.412   2.5884
+##     3   parent     71.9    69.412   2.4884
+##     5   parent     50.3    57.330  -7.0301
+##     5   parent     59.4    57.330   2.0699
+##     7   parent     47.0    47.352  -0.3515
+##     7   parent     45.1    47.352  -2.2515
+##    14   parent     27.7    24.247   3.4528
+##    14   parent     27.3    24.247   3.0528
+##    21   parent     10.0    12.416  -2.4163
+##    21   parent     10.4    12.416  -2.0163
+##    30   parent      2.9     5.251  -2.3513
+##    30   parent      4.0     5.251  -1.2513

A plot of the fit is obtained with the plot function for mkinfit objects.

-plot(m.L1.SFO, show_errmin = TRUE, main = "FOCUS L1 - SFO")

+plot(m.L1.SFO, show_errmin = TRUE, main = "FOCUS L1 - SFO")

The residual plot can be easily obtained by

-mkinresplot(m.L1.SFO, ylab = "Observed", xlab = "Time")

+mkinresplot(m.L1.SFO, ylab = "Observed", xlab = "Time")

For comparison, the FOMC model is fitted as well, and the $\chi^2$ error level is checked.

-m.L1.FOMC <- mkinfit("FOMC", FOCUS_2006_L1_mkin, quiet=TRUE)

## Warning in mkinfit("FOMC", FOCUS_2006_L1_mkin, quiet = TRUE): Optimisation did not converge:
-## false convergence (8)

+m.L1.FOMC <- mkinfit("FOMC", FOCUS_2006_L1_mkin, quiet=TRUE)

## Warning in mkinfit("FOMC", FOCUS_2006_L1_mkin, quiet = TRUE): Optimisation did not converge:
+## false convergence (8)

-plot(m.L1.FOMC, show_errmin = TRUE, main = "FOCUS L1 - FOMC")

+plot(m.L1.FOMC, show_errmin = TRUE, main = "FOCUS L1 - FOMC")

-summary(m.L1.FOMC, data = FALSE)

## Warning in sqrt(diag(covar)): NaNs produced

## Warning in sqrt(1/diag(V)): NaNs produced

## Warning in cov2cor(ans$covar): diag(.) had 0 or NA entries; non-finite result is
-## doubtful

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:40 2021 
-## Date of summary: Mon Feb 15 17:13:40 2021 
-## 
-## Equations:
-## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 369 model solutions performed in 0.084 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0 89.85  state
-## alpha     1.00 deparm
-## beta     10.00 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##               value lower upper
-## parent_0  89.850000  -Inf   Inf
-## log_alpha  0.000000  -Inf   Inf
-## log_beta   2.302585  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## 
-## Warning(s): 
-## Optimisation did not converge:
-## false convergence (8)
-## 
-## Results:
-## 
-##        AIC      BIC   logLik
-##   95.88781 99.44929 -43.9439
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##           Estimate Std. Error  Lower  Upper
-## parent_0     92.47     1.2820 89.720 95.220
-## log_alpha    13.78        NaN    NaN    NaN
-## log_beta     16.13        NaN    NaN    NaN
-## sigma         2.78     0.4598  1.794  3.766
-## 
-## Parameter correlation:
-##            parent_0 log_alpha log_beta     sigma
-## parent_0  1.0000000       NaN      NaN 0.0001671
-## log_alpha       NaN         1      NaN       NaN
-## log_beta        NaN       NaN        1       NaN
-## sigma     0.0001671       NaN      NaN 1.0000000
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##           Estimate t value Pr(>t)  Lower  Upper
-## parent_0 9.247e+01      NA     NA 89.720 95.220
-## alpha    9.658e+05      NA     NA     NA     NA
-## beta     1.010e+07      NA     NA     NA     NA
-## sigma    2.780e+00      NA     NA  1.794  3.766
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   3.619       3  6
-## parent     3.619       3  6
-## 
-## Estimated disappearance times:
-##        DT50  DT90 DT50back
-## parent 7.25 24.08     7.25

+summary(m.L1.FOMC, data = FALSE) +

## Warning in sqrt(diag(covar)): NaNs produced

## Warning in sqrt(1/diag(V)): NaNs produced

## Warning in cov2cor(ans$covar): diag(.) had 0 or NA entries; non-finite result is
+## doubtful

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:18 2022 
+## Date of summary: Wed Aug 10 15:28:18 2022 
+## 
+## Equations:
+## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 369 model solutions performed in 0.082 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0 89.85  state
+## alpha     1.00 deparm
+## beta     10.00 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##               value lower upper
+## parent_0  89.850000  -Inf   Inf
+## log_alpha  0.000000  -Inf   Inf
+## log_beta   2.302585  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## 
+## Warning(s): 
+## Optimisation did not converge:
+## false convergence (8)
+## 
+## Results:
+## 
+##        AIC      BIC   logLik
+##   95.88781 99.44929 -43.9439
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##           Estimate Std. Error  Lower  Upper
+## parent_0     92.47     1.2820 89.720 95.220
+## log_alpha    13.78        NaN    NaN    NaN
+## log_beta     16.13        NaN    NaN    NaN
+## sigma         2.78     0.4598  1.794  3.766
+## 
+## Parameter correlation:
+##            parent_0 log_alpha log_beta     sigma
+## parent_0  1.0000000       NaN      NaN 0.0001671
+## log_alpha       NaN         1      NaN       NaN
+## log_beta        NaN       NaN        1       NaN
+## sigma     0.0001671       NaN      NaN 1.0000000
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##           Estimate t value Pr(>t)  Lower  Upper
+## parent_0 9.247e+01      NA     NA 89.720 95.220
+## alpha    9.658e+05      NA     NA     NA     NA
+## beta     1.010e+07      NA     NA     NA     NA
+## sigma    2.780e+00      NA     NA  1.794  3.766
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   3.619       3  6
+## parent     3.619       3  6
+## 
+## Estimated disappearance times:
+##        DT50  DT90 DT50back
+## parent 7.25 24.08     7.25

We get a warning that the default optimisation algorithm Port did not converge, which is an indication that the model is overparameterised, i.e. contains too many parameters that are ill-defined as a consequence.

And in fact, due to the higher number of parameters, and the lower number of degrees of freedom of the fit, the $\chi^2$ error level is actually higher for the FOMC model (3.6%) than for the SFO model (3.4%). Additionally, the parameters log_alpha and log_beta internally fitted in the model have excessive confidence intervals, that span more than 25 orders of magnitude (!) when backtransformed to the scale of alpha and beta. Also, the t-test for significant difference from zero does not indicate such a significant difference, with p-values greater than 0.1, and finally, the parameter correlation of log_alpha and log_beta is 1.000, clearly indicating that the model is overparameterised.

The $\chi^2$ error levels reported in Appendix 3 and Appendix 7 to the FOCUS kinetics report are rounded to integer percentages and partly deviate by one percentage point from the results calculated by mkin. The reason for this is not known. However, mkin gives the same $\chi^2$ error levels as the kinfit package and the calculation routines of the kinfit package have been extensively compared to the results obtained by the KinGUI software, as documented in the kinfit package vignette. KinGUI was the first widely used standard package in this field. Also, the calculation of $\chi^2$ error levels was compared with KinGUII, CAKE and DegKin manager in a project sponsored by the German Umweltbundesamt (Ranke 2014).

-Laboratory Data L2

Laboratory Data L2 +

The following code defines example dataset L2 from the FOCUS kinetics report, p. 287:

-FOCUS_2006_L2 = data.frame(
-  t = rep(c(0, 1, 3, 7, 14, 28), each = 2),
-  parent = c(96.1, 91.8, 41.4, 38.7,
-             19.3, 22.3, 4.6, 4.6,
-             2.6, 1.2, 0.3, 0.6))
-FOCUS_2006_L2_mkin <- mkin_wide_to_long(FOCUS_2006_L2)

-SFO fit for L2

FOCUS_2006_L2 = data.frame(
+  t = rep(c(0, 1, 3, 7, 14, 28), each = 2),
+  parent = c(96.1, 91.8, 41.4, 38.7,
+             19.3, 22.3, 4.6, 4.6,
+             2.6, 1.2, 0.3, 0.6))
+FOCUS_2006_L2_mkin <- mkin_wide_to_long(FOCUS_2006_L2)

SFO fit for L2 +

Again, the SFO model is fitted and the result is plotted. The residual plot can be obtained simply by adding the argument show_residuals to the plot command.

-m.L2.SFO <- mkinfit("SFO", FOCUS_2006_L2_mkin, quiet=TRUE)
-plot(m.L2.SFO, show_residuals = TRUE, show_errmin = TRUE,
-     main = "FOCUS L2 - SFO")

m.L2.SFO <- mkinfit("SFO", FOCUS_2006_L2_mkin, quiet=TRUE)
+plot(m.L2.SFO, show_residuals = TRUE, show_errmin = TRUE,
+     main = "FOCUS L2 - SFO")

The $\chi^2$ error level of 14% suggests that the model does not fit very well. This is also obvious from the plots of the fit, in which we have included the residual plot.

In the FOCUS kinetics report, it is stated that there is no apparent systematic error observed from the residual plot up to the measured DT90 (approximately at day 5), and there is an underestimation beyond that point.

We may add that it is difficult to judge the random nature of the residuals just from the three samplings at days 0, 1 and 3. Also, it is not clear a priori why a consistent underestimation after the approximate DT90 should be irrelevant. However, this can be rationalised by the fact that the FOCUS fate models generally only implement SFO kinetics.

-FOMC fit for L2

FOMC fit for L2 +

For comparison, the FOMC model is fitted as well, and the $\chi^2$ error level is checked.

-m.L2.FOMC <- mkinfit("FOMC", FOCUS_2006_L2_mkin, quiet = TRUE)
-plot(m.L2.FOMC, show_residuals = TRUE,
-     main = "FOCUS L2 - FOMC")

m.L2.FOMC <- mkinfit("FOMC", FOCUS_2006_L2_mkin, quiet = TRUE)
+plot(m.L2.FOMC, show_residuals = TRUE,
+     main = "FOCUS L2 - FOMC")

-summary(m.L2.FOMC, data = FALSE)

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:40 2021 
-## Date of summary: Mon Feb 15 17:13:40 2021 
-## 
-## Equations:
-## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 239 model solutions performed in 0.05 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0 93.95  state
-## alpha     1.00 deparm
-## beta     10.00 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##               value lower upper
-## parent_0  93.950000  -Inf   Inf
-## log_alpha  0.000000  -Inf   Inf
-## log_beta   2.302585  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC      BIC    logLik
-##   61.78966 63.72928 -26.89483
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##           Estimate Std. Error    Lower   Upper
-## parent_0   93.7700     1.6130 90.05000 97.4900
-## log_alpha   0.3180     0.1559 -0.04149  0.6776
-## log_beta    0.2102     0.2493 -0.36460  0.7850
-## sigma       2.2760     0.4645  1.20500  3.3470
-## 
-## Parameter correlation:
-##             parent_0  log_alpha   log_beta      sigma
-## parent_0   1.000e+00 -1.151e-01 -2.085e-01 -7.828e-09
-## log_alpha -1.151e-01  1.000e+00  9.741e-01 -1.602e-07
-## log_beta  -2.085e-01  9.741e-01  1.000e+00 -1.372e-07
-## sigma     -7.828e-09 -1.602e-07 -1.372e-07  1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##          Estimate t value    Pr(>t)   Lower  Upper
-## parent_0   93.770  58.120 4.267e-12 90.0500 97.490
-## alpha       1.374   6.414 1.030e-04  0.9594  1.969
-## beta        1.234   4.012 1.942e-03  0.6945  2.192
-## sigma       2.276   4.899 5.977e-04  1.2050  3.347
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   6.205       3  3
-## parent     6.205       3  3
-## 
-## Estimated disappearance times:
-##          DT50  DT90 DT50back
-## parent 0.8092 5.356    1.612

+summary(m.L2.FOMC, data = FALSE)

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:19 2022 
+## Date of summary: Wed Aug 10 15:28:19 2022 
+## 
+## Equations:
+## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 239 model solutions performed in 0.048 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0 93.95  state
+## alpha     1.00 deparm
+## beta     10.00 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##               value lower upper
+## parent_0  93.950000  -Inf   Inf
+## log_alpha  0.000000  -Inf   Inf
+## log_beta   2.302585  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC      BIC    logLik
+##   61.78966 63.72928 -26.89483
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##           Estimate Std. Error    Lower   Upper
+## parent_0   93.7700     1.6130 90.05000 97.4900
+## log_alpha   0.3180     0.1559 -0.04149  0.6776
+## log_beta    0.2102     0.2493 -0.36460  0.7850
+## sigma       2.2760     0.4645  1.20500  3.3470
+## 
+## Parameter correlation:
+##             parent_0  log_alpha   log_beta      sigma
+## parent_0   1.000e+00 -1.151e-01 -2.085e-01 -7.828e-09
+## log_alpha -1.151e-01  1.000e+00  9.741e-01 -1.602e-07
+## log_beta  -2.085e-01  9.741e-01  1.000e+00 -1.372e-07
+## sigma     -7.828e-09 -1.602e-07 -1.372e-07  1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##          Estimate t value    Pr(>t)   Lower  Upper
+## parent_0   93.770  58.120 4.267e-12 90.0500 97.490
+## alpha       1.374   6.414 1.030e-04  0.9594  1.969
+## beta        1.234   4.012 1.942e-03  0.6945  2.192
+## sigma       2.276   4.899 5.977e-04  1.2050  3.347
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   6.205       3  3
+## parent     6.205       3  3
+## 
+## Estimated disappearance times:
+##          DT50  DT90 DT50back
+## parent 0.8092 5.356    1.612

The error level at which the $\chi^2$ test passes is much lower in this case. Therefore, the FOMC model provides a better description of the data, as less experimental error has to be assumed in order to explain the data.

-DFOP fit for L2

DFOP fit for L2 +

Fitting the four parameter DFOP model further reduces the $\chi^2$ error level.

-m.L2.DFOP <- mkinfit("DFOP", FOCUS_2006_L2_mkin, quiet = TRUE)
-plot(m.L2.DFOP, show_residuals = TRUE, show_errmin = TRUE,
-     main = "FOCUS L2 - DFOP")

m.L2.DFOP <- mkinfit("DFOP", FOCUS_2006_L2_mkin, quiet = TRUE)
+plot(m.L2.DFOP, show_residuals = TRUE, show_errmin = TRUE,
+     main = "FOCUS L2 - DFOP")

-summary(m.L2.DFOP, data = FALSE)

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:41 2021 
-## Date of summary: Mon Feb 15 17:13:41 2021 
-## 
-## Equations:
-## d_parent/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
-##            time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
-##            * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 581 model solutions performed in 0.134 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0 93.95  state
-## k1        0.10 deparm
-## k2        0.01 deparm
-## g         0.50 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##              value lower upper
-## parent_0 93.950000  -Inf   Inf
-## log_k1   -2.302585  -Inf   Inf
-## log_k2   -4.605170  -Inf   Inf
-## g_qlogis  0.000000  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC      BIC    logLik
-##   52.36695 54.79148 -21.18347
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##          Estimate Std. Error      Lower     Upper
-## parent_0   93.950  9.998e-01    91.5900   96.3100
-## log_k1      3.112  1.842e+03 -4353.0000 4359.0000
-## log_k2     -1.088  6.285e-02    -1.2370   -0.9394
-## g_qlogis   -0.399  9.946e-02    -0.6342   -0.1638
-## sigma       1.414  2.886e-01     0.7314    2.0960
-## 
-## Parameter correlation:
-##            parent_0     log_k1     log_k2   g_qlogis      sigma
-## parent_0  1.000e+00  6.783e-07 -3.390e-10  2.665e-01 -2.967e-10
-## log_k1    6.783e-07  1.000e+00  1.116e-04 -2.196e-04 -1.031e-05
-## log_k2   -3.390e-10  1.116e-04  1.000e+00 -7.903e-01  2.917e-09
-## g_qlogis  2.665e-01 -2.196e-04 -7.903e-01  1.000e+00 -4.408e-09
-## sigma    -2.967e-10 -1.031e-05  2.917e-09 -4.408e-09  1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##          Estimate   t value    Pr(>t)   Lower   Upper
-## parent_0  93.9500 9.397e+01 2.036e-12 91.5900 96.3100
-## k1        22.4800 5.553e-04 4.998e-01  0.0000     Inf
-## k2         0.3369 1.591e+01 4.697e-07  0.2904  0.3909
-## g          0.4016 1.680e+01 3.238e-07  0.3466  0.4591
-## sigma      1.4140 4.899e+00 8.776e-04  0.7314  2.0960
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data    2.53       4  2
-## parent      2.53       4  2
-## 
-## Estimated disappearance times:
-##          DT50  DT90 DT50back DT50_k1 DT50_k2
-## parent 0.5335 5.311    1.599 0.03084   2.058

Here, the DFOP model is clearly the best-fit model for dataset L2 based on the chi^2 error level criterion. However, the failure to calculate the covariance matrix indicates that the parameter estimates correlate excessively. Therefore, the FOMC model may be preferred for this dataset.

+summary(m.L2.DFOP, data = FALSE)

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:19 2022 
+## Date of summary: Wed Aug 10 15:28:19 2022 
+## 
+## Equations:
+## d_parent/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
+##            time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
+##            * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 581 model solutions performed in 0.132 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0 93.95  state
+## k1        0.10 deparm
+## k2        0.01 deparm
+## g         0.50 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##              value lower upper
+## parent_0 93.950000  -Inf   Inf
+## log_k1   -2.302585  -Inf   Inf
+## log_k2   -4.605170  -Inf   Inf
+## g_qlogis  0.000000  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC      BIC    logLik
+##   52.36695 54.79148 -21.18347
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##          Estimate Std. Error      Lower     Upper
+## parent_0   93.950  9.998e-01    91.5900   96.3100
+## log_k1      3.112  1.842e+03 -4353.0000 4359.0000
+## log_k2     -1.088  6.285e-02    -1.2370   -0.9394
+## g_qlogis   -0.399  9.946e-02    -0.6342   -0.1638
+## sigma       1.414  2.886e-01     0.7314    2.0960
+## 
+## Parameter correlation:
+##            parent_0     log_k1     log_k2   g_qlogis      sigma
+## parent_0  1.000e+00  6.783e-07 -3.390e-10  2.665e-01 -2.967e-10
+## log_k1    6.783e-07  1.000e+00  1.116e-04 -2.196e-04 -1.031e-05
+## log_k2   -3.390e-10  1.116e-04  1.000e+00 -7.903e-01  2.917e-09
+## g_qlogis  2.665e-01 -2.196e-04 -7.903e-01  1.000e+00 -4.408e-09
+## sigma    -2.967e-10 -1.031e-05  2.917e-09 -4.408e-09  1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##          Estimate   t value    Pr(>t)   Lower   Upper
+## parent_0  93.9500 9.397e+01 2.036e-12 91.5900 96.3100
+## k1        22.4800 5.553e-04 4.998e-01  0.0000     Inf
+## k2         0.3369 1.591e+01 4.697e-07  0.2904  0.3909
+## g          0.4016 1.680e+01 3.238e-07  0.3466  0.4591
+## sigma      1.4140 4.899e+00 8.776e-04  0.7314  2.0960
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data    2.53       4  2
+## parent      2.53       4  2
+## 
+## Estimated disappearance times:
+##          DT50  DT90 DT50back DT50_k1 DT50_k2
+## parent 0.5335 5.311    1.599 0.03084   2.058

Here, the DFOP model is clearly the best-fit model for dataset L2 based on the chi^2 error level criterion.

-Laboratory Data L3

Laboratory Data L3 +

The following code defines example dataset L3 from the FOCUS kinetics report, p. 290.

-FOCUS_2006_L3 = data.frame(
-  t = c(0, 3, 7, 14, 30, 60, 91, 120),
-  parent = c(97.8, 60, 51, 43, 35, 22, 15, 12))
-FOCUS_2006_L3_mkin <- mkin_wide_to_long(FOCUS_2006_L3)

-Fit multiple models

FOCUS_2006_L3 = data.frame(
+  t = c(0, 3, 7, 14, 30, 60, 91, 120),
+  parent = c(97.8, 60, 51, 43, 35, 22, 15, 12))
+FOCUS_2006_L3_mkin <- mkin_wide_to_long(FOCUS_2006_L3)

Fit multiple models +

As of mkin version 0.9-39 (June 2015), we can fit several models to one or more datasets in one call to the function mmkin. The datasets have to be passed in a list, in this case a named list holding only the L3 dataset prepared above.

-# Only use one core here, not to offend the CRAN checks
-mm.L3 <- mmkin(c("SFO", "FOMC", "DFOP"), cores = 1,
-               list("FOCUS L3" = FOCUS_2006_L3_mkin), quiet = TRUE)
-plot(mm.L3)

# Only use one core here, not to offend the CRAN checks
+mm.L3 <- mmkin(c("SFO", "FOMC", "DFOP"), cores = 1,
+               list("FOCUS L3" = FOCUS_2006_L3_mkin), quiet = TRUE)
+plot(mm.L3)

The $\chi^2$ error level of 21% as well as the plot suggest that the SFO model does not fit very well. The FOMC model performs better, with an error level at which the $\chi^2$ test passes of 7%. Fitting the four parameter DFOP model further reduces the $\chi^2$ error level considerably.

-Accessing mmkin objects

Accessing mmkin objects +

The objects returned by mmkin are arranged like a matrix, with models as a row index and datasets as a column index.

We can extract the summary and plot for e.g. the DFOP fit, using square brackets for indexing which will result in the use of the summary and plot functions working on mkinfit objects.

-summary(mm.L3[["DFOP", 1]])

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:41 2021 
-## Date of summary: Mon Feb 15 17:13:42 2021 
-## 
-## Equations:
-## d_parent/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
-##            time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
-##            * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 376 model solutions performed in 0.082 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0 97.80  state
-## k1        0.10 deparm
-## k2        0.01 deparm
-## g         0.50 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##              value lower upper
-## parent_0 97.800000  -Inf   Inf
-## log_k1   -2.302585  -Inf   Inf
-## log_k2   -4.605170  -Inf   Inf
-## g_qlogis  0.000000  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC      BIC    logLik
-##   32.97732 33.37453 -11.48866
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##          Estimate Std. Error   Lower      Upper
-## parent_0  97.7500    1.01900 94.5000 101.000000
-## log_k1    -0.6612    0.10050 -0.9812  -0.341300
-## log_k2    -4.2860    0.04322 -4.4230  -4.148000
-## g_qlogis  -0.1739    0.05270 -0.3416  -0.006142
-## sigma      1.0170    0.25430  0.2079   1.827000
-## 
-## Parameter correlation:
-##            parent_0     log_k1     log_k2   g_qlogis      sigma
-## parent_0  1.000e+00  1.732e-01  2.282e-02  4.009e-01 -9.664e-08
-## log_k1    1.732e-01  1.000e+00  4.945e-01 -5.809e-01  7.147e-07
-## log_k2    2.282e-02  4.945e-01  1.000e+00 -6.812e-01  1.022e-06
-## g_qlogis  4.009e-01 -5.809e-01 -6.812e-01  1.000e+00 -7.926e-07
-## sigma    -9.664e-08  7.147e-07  1.022e-06 -7.926e-07  1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##          Estimate t value    Pr(>t)    Lower     Upper
-## parent_0 97.75000  95.960 1.248e-06 94.50000 101.00000
-## k1        0.51620   9.947 1.081e-03  0.37490   0.71090
-## k2        0.01376  23.140 8.840e-05  0.01199   0.01579
-## g         0.45660  34.920 2.581e-05  0.41540   0.49850
-## sigma     1.01700   4.000 1.400e-02  0.20790   1.82700
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   2.225       4  4
-## parent     2.225       4  4
-## 
-## Estimated disappearance times:
-##         DT50 DT90 DT50back DT50_k1 DT50_k2
-## parent 7.464  123    37.03   1.343   50.37
-## 
-## Data:
-##  time variable observed predicted residual
-##     0   parent     97.8     97.75  0.05396
-##     3   parent     60.0     60.45 -0.44933
-##     7   parent     51.0     49.44  1.56338
-##    14   parent     43.0     43.84 -0.83632
-##    30   parent     35.0     35.15 -0.14707
-##    60   parent     22.0     23.26 -1.25919
-##    91   parent     15.0     15.18 -0.18181
-##   120   parent     12.0     10.19  1.81395

+summary(mm.L3[["DFOP", 1]])

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:20 2022 
+## Date of summary: Wed Aug 10 15:28:20 2022 
+## 
+## Equations:
+## d_parent/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
+##            time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
+##            * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 376 model solutions performed in 0.079 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0 97.80  state
+## k1        0.10 deparm
+## k2        0.01 deparm
+## g         0.50 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##              value lower upper
+## parent_0 97.800000  -Inf   Inf
+## log_k1   -2.302585  -Inf   Inf
+## log_k2   -4.605170  -Inf   Inf
+## g_qlogis  0.000000  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC      BIC    logLik
+##   32.97732 33.37453 -11.48866
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##          Estimate Std. Error   Lower      Upper
+## parent_0  97.7500    1.01900 94.5000 101.000000
+## log_k1    -0.6612    0.10050 -0.9812  -0.341300
+## log_k2    -4.2860    0.04322 -4.4230  -4.148000
+## g_qlogis  -0.1739    0.05270 -0.3416  -0.006142
+## sigma      1.0170    0.25430  0.2079   1.827000
+## 
+## Parameter correlation:
+##            parent_0     log_k1     log_k2   g_qlogis      sigma
+## parent_0  1.000e+00  1.732e-01  2.282e-02  4.009e-01 -9.664e-08
+## log_k1    1.732e-01  1.000e+00  4.945e-01 -5.809e-01  7.147e-07
+## log_k2    2.282e-02  4.945e-01  1.000e+00 -6.812e-01  1.022e-06
+## g_qlogis  4.009e-01 -5.809e-01 -6.812e-01  1.000e+00 -7.926e-07
+## sigma    -9.664e-08  7.147e-07  1.022e-06 -7.926e-07  1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##          Estimate t value    Pr(>t)    Lower     Upper
+## parent_0 97.75000  95.960 1.248e-06 94.50000 101.00000
+## k1        0.51620   9.947 1.081e-03  0.37490   0.71090
+## k2        0.01376  23.140 8.840e-05  0.01199   0.01579
+## g         0.45660  34.920 2.581e-05  0.41540   0.49850
+## sigma     1.01700   4.000 1.400e-02  0.20790   1.82700
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   2.225       4  4
+## parent     2.225       4  4
+## 
+## Estimated disappearance times:
+##         DT50 DT90 DT50back DT50_k1 DT50_k2
+## parent 7.464  123    37.03   1.343   50.37
+## 
+## Data:
+##  time variable observed predicted residual
+##     0   parent     97.8     97.75  0.05396
+##     3   parent     60.0     60.45 -0.44933
+##     7   parent     51.0     49.44  1.56338
+##    14   parent     43.0     43.84 -0.83632
+##    30   parent     35.0     35.15 -0.14707
+##    60   parent     22.0     23.26 -1.25919
+##    91   parent     15.0     15.18 -0.18181
+##   120   parent     12.0     10.19  1.81395

-plot(mm.L3[["DFOP", 1]], show_errmin = TRUE)

+plot(mm.L3[["DFOP", 1]], show_errmin = TRUE)

Here, a look to the model plot, the confidence intervals of the parameters and the correlation matrix suggest that the parameter estimates are reliable, and the DFOP model can be used as the best-fit model based on the $\chi^2$ error level criterion for laboratory data L3.

This is also an example where the standard t-test for the parameter g_ilr is misleading, as it tests for a significant difference from zero. In this case, zero appears to be the correct value for this parameter, and the confidence interval for the backtransformed parameter g is quite narrow.

-Laboratory Data L4

Laboratory Data L4 +

The following code defines example dataset L4 from the FOCUS kinetics report, p. 293:

-FOCUS_2006_L4 = data.frame(
-  t = c(0, 3, 7, 14, 30, 60, 91, 120),
-  parent = c(96.6, 96.3, 94.3, 88.8, 74.9, 59.9, 53.5, 49.0))
-FOCUS_2006_L4_mkin <- mkin_wide_to_long(FOCUS_2006_L4)

FOCUS_2006_L4 = data.frame(
+  t = c(0, 3, 7, 14, 30, 60, 91, 120),
+  parent = c(96.6, 96.3, 94.3, 88.8, 74.9, 59.9, 53.5, 49.0))
+FOCUS_2006_L4_mkin <- mkin_wide_to_long(FOCUS_2006_L4)

Fits of the SFO and FOMC models, plots and summaries are produced below:

-# Only use one core here, not to offend the CRAN checks
-mm.L4 <- mmkin(c("SFO", "FOMC"), cores = 1,
-               list("FOCUS L4" = FOCUS_2006_L4_mkin),
-               quiet = TRUE)
-plot(mm.L4)

# Only use one core here, not to offend the CRAN checks
+mm.L4 <- mmkin(c("SFO", "FOMC"), cores = 1,
+               list("FOCUS L4" = FOCUS_2006_L4_mkin),
+               quiet = TRUE)
+plot(mm.L4)

The $\chi^2$ error level of 3.3% as well as the plot suggest that the SFO model fits very well. The error level at which the $\chi^2$ test passes is slightly lower for the FOMC model. However, the difference appears negligible.

-summary(mm.L4[["SFO", 1]], data = FALSE)

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:42 2021 
-## Date of summary: Mon Feb 15 17:13:42 2021 
-## 
-## Equations:
-## d_parent/dt = - k_parent * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 142 model solutions performed in 0.03 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0  96.6  state
-## k_parent   0.1 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##                  value lower upper
-## parent_0     96.600000  -Inf   Inf
-## log_k_parent -2.302585  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC      BIC    logLik
-##   47.12133 47.35966 -20.56067
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##              Estimate Std. Error  Lower   Upper
-## parent_0       96.440    1.69900 92.070 100.800
-## log_k_parent   -5.030    0.07059 -5.211  -4.848
-## sigma           3.162    0.79050  1.130   5.194
-## 
-## Parameter correlation:
-##               parent_0 log_k_parent     sigma
-## parent_0     1.000e+00    5.938e-01 3.387e-07
-## log_k_parent 5.938e-01    1.000e+00 5.830e-07
-## sigma        3.387e-07    5.830e-07 1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##           Estimate t value    Pr(>t)     Lower     Upper
-## parent_0 96.440000   56.77 1.604e-08 92.070000 1.008e+02
-## k_parent  0.006541   14.17 1.578e-05  0.005455 7.842e-03
-## sigma     3.162000    4.00 5.162e-03  1.130000 5.194e+00
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   3.287       2  6
-## parent     3.287       2  6
-## 
-## Estimated disappearance times:
-##        DT50 DT90
-## parent  106  352

+summary(mm.L4[["SFO", 1]], data = FALSE) +

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:21 2022 
+## Date of summary: Wed Aug 10 15:28:21 2022 
+## 
+## Equations:
+## d_parent/dt = - k_parent * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 142 model solutions performed in 0.03 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0  96.6  state
+## k_parent   0.1 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##                  value lower upper
+## parent_0     96.600000  -Inf   Inf
+## log_k_parent -2.302585  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC      BIC    logLik
+##   47.12133 47.35966 -20.56067
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##              Estimate Std. Error  Lower   Upper
+## parent_0       96.440    1.69900 92.070 100.800
+## log_k_parent   -5.030    0.07059 -5.211  -4.848
+## sigma           3.162    0.79050  1.130   5.194
+## 
+## Parameter correlation:
+##               parent_0 log_k_parent     sigma
+## parent_0     1.000e+00    5.938e-01 3.387e-07
+## log_k_parent 5.938e-01    1.000e+00 5.830e-07
+## sigma        3.387e-07    5.830e-07 1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##           Estimate t value    Pr(>t)     Lower     Upper
+## parent_0 96.440000   56.77 1.604e-08 92.070000 1.008e+02
+## k_parent  0.006541   14.17 1.578e-05  0.005455 7.842e-03
+## sigma     3.162000    4.00 5.162e-03  1.130000 5.194e+00
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   3.287       2  6
+## parent     3.287       2  6
+## 
+## Estimated disappearance times:
+##        DT50 DT90
+## parent  106  352

-summary(mm.L4[["FOMC", 1]], data = FALSE)

## mkin version used for fitting:    1.0.3.9000 
-## R version used for fitting:       4.0.3 
-## Date of fit:     Mon Feb 15 17:13:42 2021 
-## Date of summary: Mon Feb 15 17:13:42 2021 
-## 
-## Equations:
-## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
-## 
-## Model predictions using solution type analytical 
-## 
-## Fitted using 224 model solutions performed in 0.046 s
-## 
-## Error model: Constant variance 
-## 
-## Error model algorithm: OLS 
-## 
-## Starting values for parameters to be optimised:
-##          value   type
-## parent_0  96.6  state
-## alpha      1.0 deparm
-## beta      10.0 deparm
-## 
-## Starting values for the transformed parameters actually optimised:
-##               value lower upper
-## parent_0  96.600000  -Inf   Inf
-## log_alpha  0.000000  -Inf   Inf
-## log_beta   2.302585  -Inf   Inf
-## 
-## Fixed parameter values:
-## None
-## 
-## Results:
-## 
-##        AIC      BIC    logLik
-##   40.37255 40.69032 -16.18628
-## 
-## Optimised, transformed parameters with symmetric confidence intervals:
-##           Estimate Std. Error   Lower    Upper
-## parent_0   99.1400     1.2670 95.6300 102.7000
-## log_alpha  -0.3506     0.2616 -1.0770   0.3756
-## log_beta    4.1740     0.3938  3.0810   5.2670
-## sigma       1.8300     0.4575  0.5598   3.1000
-## 
-## Parameter correlation:
-##             parent_0  log_alpha   log_beta      sigma
-## parent_0   1.000e+00 -4.696e-01 -5.543e-01 -2.468e-07
-## log_alpha -4.696e-01  1.000e+00  9.889e-01  2.478e-08
-## log_beta  -5.543e-01  9.889e-01  1.000e+00  5.211e-08
-## sigma     -2.468e-07  2.478e-08  5.211e-08  1.000e+00
-## 
-## Backtransformed parameters:
-## Confidence intervals for internally transformed parameters are asymmetric.
-## t-test (unrealistically) based on the assumption of normal distribution
-## for estimators of untransformed parameters.
-##          Estimate t value    Pr(>t)   Lower   Upper
-## parent_0  99.1400  78.250 7.993e-08 95.6300 102.700
-## alpha      0.7042   3.823 9.365e-03  0.3407   1.456
-## beta      64.9800   2.540 3.201e-02 21.7800 193.900
-## sigma      1.8300   4.000 8.065e-03  0.5598   3.100
-## 
-## FOCUS Chi2 error levels in percent:
-##          err.min n.optim df
-## All data   2.029       3  5
-## parent     2.029       3  5
-## 
-## Estimated disappearance times:
-##         DT50 DT90 DT50back
-## parent 108.9 1644    494.9

+summary(mm.L4[["FOMC", 1]], data = FALSE) +

## mkin version used for fitting:    1.1.2 
+## R version used for fitting:       4.2.1 
+## Date of fit:     Wed Aug 10 15:28:21 2022 
+## Date of summary: Wed Aug 10 15:28:21 2022 
+## 
+## Equations:
+## d_parent/dt = - (alpha/beta) * 1/((time/beta) + 1) * parent
+## 
+## Model predictions using solution type analytical 
+## 
+## Fitted using 224 model solutions performed in 0.045 s
+## 
+## Error model: Constant variance 
+## 
+## Error model algorithm: OLS 
+## 
+## Starting values for parameters to be optimised:
+##          value   type
+## parent_0  96.6  state
+## alpha      1.0 deparm
+## beta      10.0 deparm
+## 
+## Starting values for the transformed parameters actually optimised:
+##               value lower upper
+## parent_0  96.600000  -Inf   Inf
+## log_alpha  0.000000  -Inf   Inf
+## log_beta   2.302585  -Inf   Inf
+## 
+## Fixed parameter values:
+## None
+## 
+## Results:
+## 
+##        AIC      BIC    logLik
+##   40.37255 40.69032 -16.18628
+## 
+## Optimised, transformed parameters with symmetric confidence intervals:
+##           Estimate Std. Error   Lower    Upper
+## parent_0   99.1400     1.2670 95.6300 102.7000
+## log_alpha  -0.3506     0.2616 -1.0770   0.3756
+## log_beta    4.1740     0.3938  3.0810   5.2670
+## sigma       1.8300     0.4575  0.5598   3.1000
+## 
+## Parameter correlation:
+##             parent_0  log_alpha   log_beta      sigma
+## parent_0   1.000e+00 -4.696e-01 -5.543e-01 -2.468e-07
+## log_alpha -4.696e-01  1.000e+00  9.889e-01  2.478e-08
+## log_beta  -5.543e-01  9.889e-01  1.000e+00  5.211e-08
+## sigma     -2.468e-07  2.478e-08  5.211e-08  1.000e+00
+## 
+## Backtransformed parameters:
+## Confidence intervals for internally transformed parameters are asymmetric.
+## t-test (unrealistically) based on the assumption of normal distribution
+## for estimators of untransformed parameters.
+##          Estimate t value    Pr(>t)   Lower   Upper
+## parent_0  99.1400  78.250 7.993e-08 95.6300 102.700
+## alpha      0.7042   3.823 9.365e-03  0.3407   1.456
+## beta      64.9800   2.540 3.201e-02 21.7800 193.900
+## sigma      1.8300   4.000 8.065e-03  0.5598   3.100
+## 
+## FOCUS Chi2 error levels in percent:
+##          err.min n.optim df
+## All data   2.029       3  5
+## parent     2.029       3  5
+## 
+## Estimated disappearance times:
+##         DT50 DT90 DT50back
+## parent 108.9 1644    494.9

-References

References +

Ranke, Johannes. 2014. “Prüfung und Validierung von Modellierungssoftware als Alternative zu ModelMaker 4.0.” Umweltbundesamt Projektnummer 27452.

@@ -801,11 +806,13 @@ @@ -814,5 +821,7 @@ + + diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-10-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-10-1.png index e9c0b0a0..b2bff18f 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-10-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-10-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-12-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-12-1.png index 3e03954d..d613c035 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-12-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-12-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-13-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-13-1.png index 8c9e8fd4..8387a272 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-13-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-13-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-15-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-15-1.png index b3aa8334..74f0fc48 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-15-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-15-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-4-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-4-1.png index 477829a5..1c56cb20 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-4-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-4-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-5-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-5-1.png index e8f21107..4247131e 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-5-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-5-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-6-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-6-1.png index c0e08884..b6130527 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-6-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-6-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-8-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-8-1.png index 310b4f3b..dea51d58 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-8-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-8-1.png differ diff --git a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-9-1.png b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-9-1.png index 570f0026..05460304 100644 Binary files a/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-9-1.png and b/docs/dev/articles/FOCUS_L_files/figure-html/unnamed-chunk-9-1.png differ diff --git a/docs/dev/articles/index.html b/docs/dev/articles/index.html index d1ca6668..526708a4 100644 --- a/docs/dev/articles/index.html +++ b/docs/dev/articles/index.html @@ -17,7 +17,7 @@

@@ -26,7 +26,7 @@ Functions and data

Example evaluation of FOCUS Laboratory Data L1 to L3

+ Example evaluations of dimethenamid data from 2018 with nonlinear mixed-effects models +

Example evaluation of FOCUS Example Dataset Z

@@ -109,7 +112,7 @@

Site built with pkgdown 2.0.2.

Site built with pkgdown 2.0.6.

diff --git a/docs/dev/articles/web_only/benchmarks.html b/docs/dev/articles/web_only/benchmarks.html index a6d52649..3dbf2881 100644 --- a/docs/dev/articles/web_only/benchmarks.html +++ b/docs/dev/articles/web_only/benchmarks.html @@ -20,6 +20,8 @@ + +

Each system is characterized by its CPU type, the operating system type and the mkin version. Currently only values for one system are available. A compiler was available, so if no analytical solution was available, compiled ODE models are used.

-Test cases

Parent only:

Each system is characterized by the operating system type, the CPU type, the mkin version, and, as in June 2022 the current R version lead to worse performance, the R version. A compiler was available, so if no analytical solution was available, compiled ODE models are used.

Every fit is only performed once, so the accuracy of the benchmarks is limited.

The following wrapper function for mmkin is used because the way the error model is specified was changed in mkin version 0.9.49.1.

-FOCUS_C <- FOCUS_2006_C
-FOCUS_D <- subset(FOCUS_2006_D, value != 0)
-parent_datasets <- list(FOCUS_C, FOCUS_D)
-
-t1 <- system.time(mmkin_bench(c("SFO", "FOMC", "DFOP", "HS"), parent_datasets))[["elapsed"]]
-t2 <- system.time(mmkin_bench(c("SFO", "FOMC", "DFOP", "HS"), parent_datasets,
-    error_model = "tc"))[["elapsed"]]

One metabolite:

if (packageVersion("mkin") > "0.9.48.1") {
+  mmkin_bench <- function(models, datasets, error_model = "const") {
+    mmkin(models, datasets, error_model = error_model, cores = 1, quiet = TRUE)
+  }
+} else {
+  mmkin_bench <- function(models, datasets, error_model = NULL) {
+    mmkin(models, datasets, reweight.method = error_model, cores = 1, quiet = TRUE)
+  }
+}

Test cases +

Parent only:

-SFO_SFO <- mkinmod(
-  parent = mkinsub("SFO", "m1"),
-  m1 = mkinsub("SFO"))
-FOMC_SFO <- mkinmod(
-  parent = mkinsub("FOMC", "m1"),
-  m1 = mkinsub("SFO"))
-DFOP_SFO <- mkinmod(
-  parent = mkinsub("FOMC", "m1"),
-  m1 = mkinsub("SFO"))
-t3 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D)))[["elapsed"]]
-t4 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D),
-    error_model = "tc"))[["elapsed"]]
-t5 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D),
-    error_model = "obs"))[["elapsed"]]

Two metabolites, synthetic data:

FOCUS_C <- FOCUS_2006_C
+FOCUS_D <- subset(FOCUS_2006_D, value != 0)
+parent_datasets <- list(FOCUS_C, FOCUS_D)
+
+
+t1 <- system.time(mmkin_bench(c("SFO", "FOMC", "DFOP", "HS"), parent_datasets))[["elapsed"]]
+t2 <- system.time(mmkin_bench(c("SFO", "FOMC", "DFOP", "HS"), parent_datasets,
+    error_model = "tc"))[["elapsed"]]

One metabolite:

-m_synth_SFO_lin <- mkinmod(parent = mkinsub("SFO", "M1"),
-                           M1 = mkinsub("SFO", "M2"),
-                           M2 = mkinsub("SFO"),
-                           use_of_ff = "max", quiet = TRUE)
-
-m_synth_DFOP_par <- mkinmod(parent = mkinsub("DFOP", c("M1", "M2")),
-                           M1 = mkinsub("SFO"),
-                           M2 = mkinsub("SFO"),
-                           use_of_ff = "max", quiet = TRUE)
-
-SFO_lin_a <- synthetic_data_for_UBA_2014[[1]]$data
-
-DFOP_par_c <- synthetic_data_for_UBA_2014[[12]]$data
-
-t6 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a)))[["elapsed"]]
-t7 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c)))[["elapsed"]]
-
-t8 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a),
-    error_model = "tc"))[["elapsed"]]
-t9 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c),
-    error_model = "tc"))[["elapsed"]]
-
-t10 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a),
-    error_model = "obs"))[["elapsed"]]
-t11 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c),
-    error_model = "obs"))[["elapsed"]]

SFO_SFO <- mkinmod(
+  parent = mkinsub("SFO", "m1"),
+  m1 = mkinsub("SFO"))
+FOMC_SFO <- mkinmod(
+  parent = mkinsub("FOMC", "m1"),
+  m1 = mkinsub("SFO"))
+DFOP_SFO <- mkinmod(
+  parent = mkinsub("FOMC", "m1"),
+  m1 = mkinsub("SFO"))
+t3 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D)))[["elapsed"]]
+t4 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D),
+    error_model = "tc"))[["elapsed"]]
+t5 <- system.time(mmkin_bench(list(SFO_SFO, FOMC_SFO, DFOP_SFO), list(FOCUS_D),
+    error_model = "obs"))[["elapsed"]]

Two metabolites, synthetic data:

-mkin_benchmarks[system_string, paste0("t", 1:11)] <-
-  c(t1, t2, t3, t4, t5, t6, t7, t8, t9, t10, t11)
-save(mkin_benchmarks, file = "~/git/mkin/vignettes/web_only/mkin_benchmarks.rda")

m_synth_SFO_lin <- mkinmod(parent = mkinsub("SFO", "M1"),
+                           M1 = mkinsub("SFO", "M2"),
+                           M2 = mkinsub("SFO"),
+                           use_of_ff = "max", quiet = TRUE)
+
+m_synth_DFOP_par <- mkinmod(parent = mkinsub("DFOP", c("M1", "M2")),
+                           M1 = mkinsub("SFO"),
+                           M2 = mkinsub("SFO"),
+                           use_of_ff = "max", quiet = TRUE)
+
+SFO_lin_a <- synthetic_data_for_UBA_2014[[1]]$data
+
+DFOP_par_c <- synthetic_data_for_UBA_2014[[12]]$data
+
+t6 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a)))[["elapsed"]]
+t7 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c)))[["elapsed"]]
+
+t8 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a),
+    error_model = "tc"))[["elapsed"]]
+t9 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c),
+    error_model = "tc"))[["elapsed"]]
+
+t10 <- system.time(mmkin_bench(list(m_synth_SFO_lin), list(SFO_lin_a),
+    error_model = "obs"))[["elapsed"]]
+t11 <- system.time(mmkin_bench(list(m_synth_DFOP_par), list(DFOP_par_c),
+    error_model = "obs"))[["elapsed"]]

-Results

Currently, we only have benchmark information on one system, therefore only the mkin version is shown with the results below. Timings are in seconds, shorter is better. All results were obtained by serial, i.e. not using multiple computing cores.

Benchmarks for all available error models are shown.

-Parent only

Results +

Benchmarks for all available error models are shown. They are intended for improving mkin, not for comparing CPUs or operating systems. All trademarks belong to their respective owners.

Parent only +

Constant variance (t1) and two-component error model (t2) for four models fitted to two datasets, i.e. eight fits for each test.

- - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + - - + + - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

mkin version	t1 [s]	t2 [s]	OS	CPU	R	mkin	t1	t2
Linux	Ryzen 7 1700	NA	0.9.48.1	3.610	11.019
Linux	Ryzen 7 1700	NA	0.9.49.1	8.184	22.889
Linux	Ryzen 7 1700	NA	0.9.49.2	7.064	12.558
Linux	Ryzen 7 1700	NA	0.9.49.3	7.296	21.239
Linux	Ryzen 7 1700	NA	0.9.49.4	5.936	20.545
Linux	Ryzen 7 1700	NA	0.9.50.2	1.714	3.971
Linux	Ryzen 7 1700	NA	0.9.50.3	1.752	4.156
Linux	Ryzen 7 1700	NA	0.9.50.4	1.786	3.729
Linux	Ryzen 7 1700	NA	1.0.3	1.722	3.419	1.881	3.504
1.0.3.9000	2.770	3.458	Linux	Ryzen 7 1700	NA	1.0.4	1.867	3.450
Linux	Ryzen 7 1700	4.1.3	1.1.0	1.791	3.289
Linux	Ryzen 7 1700	4.2.1	1.1.0	1.842	3.453
Linux	i7-4710MQ	4.2.1	1.1.0	1.959	4.116
Linux	i7-4710MQ	4.1.3	1.1.0	1.877	3.906
Linux	i7-4710MQ	4.2.1	1.1.1	1.644	3.172
Linux	Ryzen 7 1700	4.2.1	1.1.1	1.770	3.377
Linux	Ryzen 7 1700	4.2.1	1.1.2	1.940	3.619

-One metabolite

One metabolite +

Constant variance (t3), two-component error model (t4), and variance by variable (t5) for three models fitted to one dataset, i.e. three fits for each test.

- - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + - - - + + + - - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + +

mkin version	t3 [s]	t4 [s]	t5 [s]	OS	CPU	R	mkin	t3	t4	t5
Linux	Ryzen 7 1700	NA	0.9.48.1	3.764	14.347	9.495
Linux	Ryzen 7 1700	NA	0.9.49.1	4.649	13.789	6.395
Linux	Ryzen 7 1700	NA	0.9.49.2	4.786	8.461	5.675
Linux	Ryzen 7 1700	NA	0.9.49.3	4.510	13.805	7.386
Linux	Ryzen 7 1700	NA	0.9.49.4	4.446	15.335	6.002
Linux	Ryzen 7 1700	NA	0.9.50.2	1.402	6.174	2.764
Linux	Ryzen 7 1700	NA	0.9.50.3	1.430	6.615	2.878
Linux	Ryzen 7 1700	NA	0.9.50.4	1.397	7.251	2.810
Linux	Ryzen 7 1700	NA	1.0.3	1.402	6.343	2.802	1.430	6.344	2.798
1.0.3.9000	1.405	6.417	2.824	Linux	Ryzen 7 1700	NA	1.0.4	1.415	6.364	2.820
Linux	Ryzen 7 1700	4.1.3	1.1.0	1.310	6.279	2.681
Linux	Ryzen 7 1700	4.2.1	1.1.0	3.802	21.247	8.461
Linux	i7-4710MQ	4.2.1	1.1.0	3.334	19.521	7.565
Linux	i7-4710MQ	4.1.3	1.1.0	1.578	8.058	3.339
Linux	i7-4710MQ	4.2.1	1.1.1	1.230	5.839	2.444
Linux	Ryzen 7 1700	4.2.1	1.1.1	1.308	5.758	2.558
Linux	Ryzen 7 1700	4.2.1	1.1.2	1.490	6.035	2.799

-Two metabolites

Two metabolites +

Constant variance (t6 and t7), two-component error model (t8 and t9), and variance by variable (t10 and t11) for one model fitted to one dataset, i.e. one fit for each test.

- - - - - - - + + + + + + + + + + + + + @@ -341,6 +545,9 @@ + + + @@ -350,6 +557,9 @@ + + + @@ -359,6 +569,9 @@ + + + @@ -368,6 +581,9 @@ + + + @@ -377,6 +593,9 @@ + + + @@ -386,6 +605,9 @@ + + + @@ -395,6 +617,9 @@ + + + @@ -404,22 +629,112 @@ + + + - - - - - - + + + + + + - - - + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + - - - + + +

mkin version	t6 [s]	t7 [s]	t8 [s]	t9 [s]	t10 [s]	t11 [s]	OS	CPU	R	mkin	t6	t7	t8	t9	t10
Linux	Ryzen 7 1700	NA	0.9.48.1	2.623	4.587	31.267
Linux	Ryzen 7 1700	NA	0.9.49.1	2.542	4.128	5.636
Linux	Ryzen 7 1700	NA	0.9.49.2	2.723	4.478	5.574
Linux	Ryzen 7 1700	NA	0.9.49.3	2.643	4.374	7.365
Linux	Ryzen 7 1700	NA	0.9.49.4	2.635	4.259	5.626
Linux	Ryzen 7 1700	NA	0.9.50.2	0.777	1.236	2.987
Linux	Ryzen 7 1700	NA	0.9.50.3	0.858	1.264	3.073
Linux	Ryzen 7 1700	NA	0.9.50.4	0.783	1.282	3.105
Linux	Ryzen 7 1700	NA	1.0.3	0.771	1.251	1.464	3.074	1.940	2.831	0.763	1.244	1.457	3.054	1.923	2.839
1.0.3.9000	0.772	1.263	Linux	Ryzen 7 1700	NA	1.0.4	0.785	1.252	1.466	3.091	1.936	2.826
Linux	Ryzen 7 1700	4.1.3	1.1.0	0.744	1.227	1.288	3.553	1.895	2.738
Linux	Ryzen 7 1700	4.2.1	1.1.0	3.018	4.165	5.036	10.844	6.623	9.722
Linux	i7-4710MQ	4.2.1	1.1.0	2.522	3.792	4.143	11.268	5.935	8.728
Linux	i7-4710MQ	4.1.3	1.1.0	0.907	1.535	1.589	4.544	2.302	3.463
Linux	i7-4710MQ	4.2.1	1.1.1	0.678	1.095	1.149	3.247	1.658	2.472
Linux	Ryzen 7 1700	4.2.1	1.1.1	0.696	1.124	1.321	2.786	1.744	2.566
Linux	Ryzen 7 1700	4.2.1	1.1.2	0.857	1.295	1.483	3.101	1.958	2.843	2.989	1.919	2.766

@@ -438,11 +753,13 @@ @@ -451,5 +768,7 @@ + + diff --git a/docs/dev/articles/web_only/dimethenamid_2018.html b/docs/dev/articles/web_only/dimethenamid_2018.html index 6b5c8c4e..81b15cb9 100644 --- a/docs/dev/articles/web_only/dimethenamid_2018.html +++ b/docs/dev/articles/web_only/dimethenamid_2018.html @@ -34,7 +34,7 @@

@@ -44,7 +44,7 @@ Functions and data

Example evaluation of FOCUS Laboratory Data L1 to L3

+ Example evaluations of dimethenamid data from 2018 with nonlinear mixed-effects models +

Example evaluation of FOCUS Example Dataset Z

@@ -103,7 +106,7 @@

Example evaluations of the dimethenamid data from 2018

Johannes Ranke

Last change 10 February 2022, built on 28 Feb 2022

Last change 1 July 2022, built on 10 Aug 2022

Source: vignettes/web_only/dimethenamid_2018.rmd

dimethenamid_2018.rmd

@@ -116,8 +119,8 @@

Introduction

During the preparation of the journal article on nonlinear mixed-effects models in degradation kinetics (Ranke et al. 2021) and the analysis of the dimethenamid degradation data analysed therein, a need for a more detailed analysis using not only nlme and saemix, but also nlmixr for fitting the mixed-effects models was identified, as many model variants do not converge when fitted with nlme, and not all relevant error models can be fitted with saemix.

This vignette is an attempt to satisfy this need.

A first analysis of the data analysed here was presented in a recent journal article on nonlinear mixed-effects models in degradation kinetics (Ranke et al. 2021). That analysis was based on the nlme package and a development version of the saemix package that was unpublished at the time. Meanwhile, version 3.0 of the saemix package is available from the CRAN repository. Also, it turned out that there was an error in the handling of the Borstel data in the mkin package at the time, leading to the duplication of a few data points from that soil. The dataset in the mkin package has been corrected, and the interface to saemix in the mkin package has been updated to use the released version.

This vignette is intended to present an up to date analysis of the data, using the corrected dataset and released versions of mkin and saemix.

Data @@ -126,17 +129,17 @@

The data are available in the mkin package. The following code (hidden by default, please use the button to the right to show it) treats the data available for the racemic mixture dimethenamid (DMTA) and its enantiomer dimethenamid-P (DMTAP) in the same way, as no difference between their degradation behaviour was identified in the EU risk assessment. The observation times of each dataset are multiplied with the corresponding normalisation factor also available in the dataset, in order to make it possible to describe all datasets with a single set of parameters.

Also, datasets observed in the same soil are merged, resulting in dimethenamid (DMTA) data from six soils.

-library(mkin, quietly = TRUE)
-dmta_ds <- lapply(1:7, function(i) {
-  ds_i <- dimethenamid_2018$ds[[i]]$data
-  ds_i[ds_i$name == "DMTAP", "name"] <-  "DMTA"
-  ds_i$time <- ds_i$time * dimethenamid_2018$f_time_norm[i]
-  ds_i
-})
-names(dmta_ds) <- sapply(dimethenamid_2018$ds, function(ds) ds$title)
-dmta_ds[["Elliot"]] <- rbind(dmta_ds[["Elliot 1"]], dmta_ds[["Elliot 2"]])
-dmta_ds[["Elliot 1"]] <- NULL
-dmta_ds[["Elliot 2"]] <- NULL

library(mkin, quietly = TRUE)
+dmta_ds <- lapply(1:7, function(i) {
+  ds_i <- dimethenamid_2018$ds[[i]]$data
+  ds_i[ds_i$name == "DMTAP", "name"] <-  "DMTA"
+  ds_i$time <- ds_i$time * dimethenamid_2018$f_time_norm[i]
+  ds_i
+})
+names(dmta_ds) <- sapply(dimethenamid_2018$ds, function(ds) ds$title)
+dmta_ds[["Elliot"]] <- rbind(dmta_ds[["Elliot 1"]], dmta_ds[["Elliot 2"]])
+dmta_ds[["Elliot 1"]] <- NULL
+dmta_ds[["Elliot 2"]] <- NULL

Parent degradation @@ -147,30 +150,30 @@

As a first step, to get a visual impression of the fit of the different models, we do separate evaluations for each soil using the mmkin function from the mkin package:

-f_parent_mkin_const <- mmkin(c("SFO", "DFOP"), dmta_ds,
-  error_model = "const", quiet = TRUE)
-f_parent_mkin_tc <- mmkin(c("SFO", "DFOP"), dmta_ds,
-  error_model = "tc", quiet = TRUE)

f_parent_mkin_const <- mmkin(c("SFO", "DFOP"), dmta_ds,
+  error_model = "const", quiet = TRUE)
+f_parent_mkin_tc <- mmkin(c("SFO", "DFOP"), dmta_ds,
+  error_model = "tc", quiet = TRUE)

The plot of the individual SFO fits shown below suggests that at least in some datasets the degradation slows down towards later time points, and that the scatter of the residuals error is smaller for smaller values (panel to the right):

-plot(mixed(f_parent_mkin_const["SFO", ]))

+plot(mixed(f_parent_mkin_const["SFO", ]))

Using biexponential decline (DFOP) results in a slightly more random scatter of the residuals:

-plot(mixed(f_parent_mkin_const["DFOP", ]))

+plot(mixed(f_parent_mkin_const["DFOP", ]))

The population curve (bold line) in the above plot results from taking the mean of the individual transformed parameters, i.e. of log k1 and log k2, as well as of the logit of the g parameter of the DFOP model). Here, this procedure does not result in parameters that represent the degradation well, because in some datasets the fitted value for k2 is extremely close to zero, leading to a log k2 value that dominates the average. This is alleviated if only rate constants that pass the t-test for significant difference from zero (on the untransformed scale) are considered in the averaging:

-plot(mixed(f_parent_mkin_const["DFOP", ]), test_log_parms = TRUE)

+plot(mixed(f_parent_mkin_const["DFOP", ]), test_log_parms = TRUE)

While this is visually much more satisfactory, such an average procedure could introduce a bias, as not all results from the individual fits enter the population curve with the same weight. This is where nonlinear mixed-effects models can help out by treating all datasets with equally by fitting a parameter distribution model together with the degradation model and the error model (see below).

The remaining trend of the residuals to be higher for higher predicted residues is reduced by using the two-component error model:

-plot(mixed(f_parent_mkin_tc["DFOP", ]), test_log_parms = TRUE)

+plot(mixed(f_parent_mkin_tc["DFOP", ]), test_log_parms = TRUE)

However, note that in the case of using this error model, the fits to the Flaach and BBA 2.3 datasets appear to be ill-defined, indicated by the fact that they did not converge:

-print(f_parent_mkin_tc["DFOP", ])

+print(f_parent_mkin_tc["DFOP", ])

<mmkin> object
 Status of individual fits:
 
@@ -178,9 +181,9 @@ Status of individual fits:
 model  Calke Borstel Flaach BBA 2.2 BBA 2.3 Elliot
   DFOP OK    OK      C      OK      C       OK    
 
-OK: No warnings
 C: Optimisation did not converge:
-iteration limit reached without convergence (10)

+iteration limit reached without convergence (10) +OK: No warnings

Nonlinear mixed-effects models @@ -191,92 +194,146 @@ iteration limit reached without convergence (10)

The nlme package was the first R extension providing facilities to fit nonlinear mixed-effects models. We would like to do model selection from all four combinations of degradation models and error models based on the AIC. However, fitting the DFOP model with constant variance and using default control parameters results in an error, signalling that the maximum number of 50 iterations was reached, potentially indicating overparameterisation. Nevertheless, the algorithm converges when the two-component error model is used in combination with the DFOP model. This can be explained by the fact that the smaller residues observed at later sampling times get more weight when using the two-component error model which will counteract the tendency of the algorithm to try parameter combinations unsuitable for fitting these data.

-library(nlme)
-f_parent_nlme_sfo_const <- nlme(f_parent_mkin_const["SFO", ])
-# f_parent_nlme_dfop_const <- nlme(f_parent_mkin_const["DFOP", ])
-f_parent_nlme_sfo_tc <- nlme(f_parent_mkin_tc["SFO", ])
-f_parent_nlme_dfop_tc <- nlme(f_parent_mkin_tc["DFOP", ])

library(nlme)
+f_parent_nlme_sfo_const <- nlme(f_parent_mkin_const["SFO", ])
+# f_parent_nlme_dfop_const <- nlme(f_parent_mkin_const["DFOP", ])
+f_parent_nlme_sfo_tc <- nlme(f_parent_mkin_tc["SFO", ])
+f_parent_nlme_dfop_tc <- nlme(f_parent_mkin_tc["DFOP", ])

Note that a certain degree of overparameterisation is also indicated by a warning obtained when fitting DFOP with the two-component error model (‘false convergence’ in the ‘LME step’ in iteration 3). However, as this warning does not occur in later iterations, and specifically not in the last of the 6 iterations, we can ignore this warning.

The model comparison function of the nlme package can directly be applied to these fits showing a much lower AIC for the DFOP model fitted with the two-component error model. Also, the likelihood ratio test indicates that this difference is significant as the p-value is below 0.0001.

-anova(
-  f_parent_nlme_sfo_const, f_parent_nlme_sfo_tc, f_parent_nlme_dfop_tc
-)

anova(
+  f_parent_nlme_sfo_const, f_parent_nlme_sfo_tc, f_parent_nlme_dfop_tc
+)

                        Model df    AIC    BIC  logLik   Test L.Ratio p-value
 f_parent_nlme_sfo_const     1  5 796.60 811.82 -393.30                       
 f_parent_nlme_sfo_tc        2  6 798.60 816.86 -393.30 1 vs 2    0.00   0.998
 f_parent_nlme_dfop_tc       3 10 671.91 702.34 -325.96 2 vs 3  134.69  <.0001

In addition to these fits, attempts were also made to include correlations between random effects by using the log Cholesky parameterisation of the matrix specifying them. The code used for these attempts can be made visible below.

-f_parent_nlme_sfo_const_logchol <- nlme(f_parent_mkin_const["SFO", ],
-  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k_DMTA ~ 1)))
-anova(f_parent_nlme_sfo_const, f_parent_nlme_sfo_const_logchol)
-f_parent_nlme_sfo_tc_logchol <- nlme(f_parent_mkin_tc["SFO", ],
-  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k_DMTA ~ 1)))
-anova(f_parent_nlme_sfo_tc, f_parent_nlme_sfo_tc_logchol)
-f_parent_nlme_dfop_tc_logchol <- nlme(f_parent_mkin_const["DFOP", ],
-  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k1 ~ 1, log_k2 ~ 1, g_qlogis ~ 1)))
-anova(f_parent_nlme_dfop_tc, f_parent_nlme_dfop_tc_logchol)

f_parent_nlme_sfo_const_logchol <- nlme(f_parent_mkin_const["SFO", ],
+  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k_DMTA ~ 1)))
+anova(f_parent_nlme_sfo_const, f_parent_nlme_sfo_const_logchol)
+f_parent_nlme_sfo_tc_logchol <- nlme(f_parent_mkin_tc["SFO", ],
+  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k_DMTA ~ 1)))
+anova(f_parent_nlme_sfo_tc, f_parent_nlme_sfo_tc_logchol)
+f_parent_nlme_dfop_tc_logchol <- nlme(f_parent_mkin_const["DFOP", ],
+  random = nlme::pdLogChol(list(DMTA_0 ~ 1, log_k1 ~ 1, log_k2 ~ 1, g_qlogis ~ 1)))
+anova(f_parent_nlme_dfop_tc, f_parent_nlme_dfop_tc_logchol)

While the SFO variants converge fast, the additional parameters introduced by this lead to convergence warnings for the DFOP model. The model comparison clearly show that adding correlations between random effects does not improve the fits.

The selected model (DFOP with two-component error) fitted to the data assuming no correlations between random effects is shown below.

-plot(f_parent_nlme_dfop_tc)

+plot(f_parent_nlme_dfop_tc)

saemix

The saemix package provided the first Open Source implementation of the Stochastic Approximation to the Expectation Maximisation (SAEM) algorithm. SAEM fits of degradation models can be conveniently performed using an interface to the saemix package available in current development versions of the mkin package.

The corresponding SAEM fits of the four combinations of degradation and error models are fitted below. As there is no convergence criterion implemented in the saemix package, the convergence plots need to be manually checked for every fit. As we will compare the SAEM implementation of saemix to the results obtained using the nlmixr package later, we define control settings that work well for all the parent data fits shown in this vignette.

The corresponding SAEM fits of the four combinations of degradation and error models are fitted below. As there is no convergence criterion implemented in the saemix package, the convergence plots need to be manually checked for every fit. We define control settings that work well for all the parent data fits shown in this vignette.

-library(saemix)
-saemix_control <- saemixControl(nbiter.saemix = c(800, 300), nb.chains = 15,
-    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)
-saemix_control_moreiter <- saemixControl(nbiter.saemix = c(1600, 300), nb.chains = 15,
-    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)
-saemix_control_10k <- saemixControl(nbiter.saemix = c(10000, 300), nb.chains = 15,
-    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)

library(saemix)
+saemix_control <- saemixControl(nbiter.saemix = c(800, 300), nb.chains = 15,
+    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)
+saemix_control_moreiter <- saemixControl(nbiter.saemix = c(1600, 300), nb.chains = 15,
+    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)
+saemix_control_10k <- saemixControl(nbiter.saemix = c(10000, 300), nb.chains = 15,
+    print = FALSE, save = FALSE, save.graphs = FALSE, displayProgress = FALSE)

The convergence plot for the SFO model using constant variance is shown below.

-f_parent_saemix_sfo_const <- mkin::saem(f_parent_mkin_const["SFO", ], quiet = TRUE,
-  control = saemix_control, transformations = "saemix")
-plot(f_parent_saemix_sfo_const$so, plot.type = "convergence")

f_parent_saemix_sfo_const <- mkin::saem(f_parent_mkin_const["SFO", ], quiet = TRUE,
+  control = saemix_control, transformations = "saemix")
+plot(f_parent_saemix_sfo_const$so, plot.type = "convergence")

Obviously the default number of iterations is sufficient to reach convergence. This can also be said for the SFO fit using the two-component error model.

Obviously the selected number of iterations is sufficient to reach convergence. This can also be said for the SFO fit using the two-component error model.

-f_parent_saemix_sfo_tc <- mkin::saem(f_parent_mkin_tc["SFO", ], quiet = TRUE,
-  control = saemix_control, transformations = "saemix")
-plot(f_parent_saemix_sfo_tc$so, plot.type = "convergence")

f_parent_saemix_sfo_tc <- mkin::saem(f_parent_mkin_tc["SFO", ], quiet = TRUE,
+  control = saemix_control, transformations = "saemix")
+plot(f_parent_saemix_sfo_tc$so, plot.type = "convergence")

When fitting the DFOP model with constant variance (see below), parameter convergence is not as unambiguous.

-f_parent_saemix_dfop_const <- mkin::saem(f_parent_mkin_const["DFOP", ], quiet = TRUE,
-  control = saemix_control, transformations = "saemix")
-plot(f_parent_saemix_dfop_const$so, plot.type = "convergence")

f_parent_saemix_dfop_const <- mkin::saem(f_parent_mkin_const["DFOP", ], quiet = TRUE,
+  control = saemix_control, transformations = "saemix")
+plot(f_parent_saemix_dfop_const$so, plot.type = "convergence")

This is improved when the DFOP model is fitted with the two-component error model. Convergence of the variance of k2 is enhanced, it remains more or less stable already after 200 iterations of the first phase.

-f_parent_saemix_dfop_tc <- mkin::saem(f_parent_mkin_tc["DFOP", ], quiet = TRUE,
-  control = saemix_control, transformations = "saemix")
-f_parent_saemix_dfop_tc_moreiter <- mkin::saem(f_parent_mkin_tc["DFOP", ], quiet = TRUE,
-  control = saemix_control_moreiter, transformations = "saemix")
-plot(f_parent_saemix_dfop_tc$so, plot.type = "convergence")

+print(f_parent_saemix_dfop_const) +

Kinetic nonlinear mixed-effects model fit by SAEM
+Structural model:
+d_DMTA/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
+           time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
+           * DMTA
+
+Data:
+155 observations of 1 variable(s) grouped in 6 datasets
+
+Likelihood computed by importance sampling
+  AIC BIC logLik
+  706 704   -344
+
+Fitted parameters:
+          estimate    lower   upper
+DMTA_0    97.99583 96.50079 99.4909
+k1         0.06377  0.03432  0.0932
+k2         0.00848  0.00444  0.0125
+g          0.95701  0.91313  1.0009
+a.1        1.82141  1.65974  1.9831
+SD.DMTA_0  1.64787  0.45779  2.8379
+SD.k1      0.57439  0.24731  0.9015
+SD.k2      0.03296 -2.50143  2.5673
+SD.g       1.10266  0.32371  1.8816

While the other parameters converge to credible values, the variance of k2 (omega2.k2) converges to a very small value. The printout of the saem.mmkin model shows that the estimated standard deviation of k2 across the population of soils (SD.k2) is ill-defined, indicating overparameterisation of this model.

When the DFOP model is fitted with the two-component error model, we also observe that the estimated variance of k2 becomes very small, while being ill-defined, as illustrated by the excessive confidence interval of SD.k2.

+f_parent_saemix_dfop_tc <- mkin::saem(f_parent_mkin_tc["DFOP", ], quiet = TRUE,
+  control = saemix_control, transformations = "saemix")
+f_parent_saemix_dfop_tc_moreiter <- mkin::saem(f_parent_mkin_tc["DFOP", ], quiet = TRUE,
+  control = saemix_control_moreiter, transformations = "saemix")
+plot(f_parent_saemix_dfop_tc$so, plot.type = "convergence")

Doubling the number of iterations in the first phase of the algorithm leads to a slightly lower likelihood, and therefore to slightly higher AIC and BIC values. With even more iterations, the algorithm stops with an error message. This is related to the variance of k2 approximating zero. This has been submitted as a bug to the saemix package, as the algorithm does not converge in this case.

+print(f_parent_saemix_dfop_tc)

Kinetic nonlinear mixed-effects model fit by SAEM
+Structural model:
+d_DMTA/dt = - ((k1 * g * exp(-k1 * time) + k2 * (1 - g) * exp(-k2 *
+           time)) / (g * exp(-k1 * time) + (1 - g) * exp(-k2 * time)))
+           * DMTA
+
+Data:
+155 observations of 1 variable(s) grouped in 6 datasets
+
+Likelihood computed by importance sampling
+  AIC BIC logLik
+  666 664   -323
+
+Fitted parameters:
+          estimate    lower    upper
+DMTA_0    98.27617  96.3088 100.2436
+k1         0.06437   0.0337   0.0950
+k2         0.00880   0.0063   0.0113
+g          0.95249   0.9100   0.9949
+a.1        1.06161   0.8625   1.2607
+b.1        0.02967   0.0226   0.0367
+SD.DMTA_0  2.06075   0.4187   3.7028
+SD.k1      0.59357   0.2561   0.9310
+SD.k2      0.00292 -10.2960  10.3019
+SD.g       1.05725   0.3808   1.7337

Doubling the number of iterations in the first phase of the algorithm leads to a slightly lower likelihood, and therefore to slightly higher AIC and BIC values. With even more iterations, the algorithm stops with an error message. This is related to the variance of k2 approximating zero and has been submitted as a bug to the saemix package, as the algorithm does not converge in this case.

An alternative way to fit DFOP in combination with the two-component error model is to use the model formulation with transformed parameters as used per default in mkin. When using this option, convergence is slower, but eventually the algorithm stops as well with the same error message.

The four combinations (SFO/const, SFO/tc, DFOP/const and DFOP/tc) and the version with increased iterations can be compared using the model comparison function of the saemix package:

-AIC_parent_saemix <- saemix::compare.saemix(
-  f_parent_saemix_sfo_const$so,
-  f_parent_saemix_sfo_tc$so,
-  f_parent_saemix_dfop_const$so,
-  f_parent_saemix_dfop_tc$so,
-  f_parent_saemix_dfop_tc_moreiter$so)

+AIC_parent_saemix <- saemix::compare.saemix(
+  f_parent_saemix_sfo_const$so,
+  f_parent_saemix_sfo_tc$so,
+  f_parent_saemix_dfop_const$so,
+  f_parent_saemix_dfop_tc$so,
+  f_parent_saemix_dfop_tc_moreiter$so)

Likelihoods calculated by importance sampling

-rownames(AIC_parent_saemix) <- c(
-  "SFO const", "SFO tc", "DFOP const", "DFOP tc", "DFOP tc more iterations")
-print(AIC_parent_saemix)

+rownames(AIC_parent_saemix) <- c(
+  "SFO const", "SFO tc", "DFOP const", "DFOP tc", "DFOP tc more iterations")
+print(AIC_parent_saemix)

                           AIC    BIC
 SFO const               796.38 795.34
 SFO tc                  798.38 797.13
@@ -284,149 +341,57 @@ DFOP const              705.75 703.88
 DFOP tc                 665.65 663.57
 DFOP tc more iterations 665.88 663.80

In order to check the influence of the likelihood calculation algorithms implemented in saemix, the likelihood from Gaussian quadrature is added to the best fit, and the AIC values obtained from the three methods are compared.

-f_parent_saemix_dfop_tc$so <-
-  saemix::llgq.saemix(f_parent_saemix_dfop_tc$so)
-AIC_parent_saemix_methods <- c(
-  is = AIC(f_parent_saemix_dfop_tc$so, method = "is"),
-  gq = AIC(f_parent_saemix_dfop_tc$so, method = "gq"),
-  lin = AIC(f_parent_saemix_dfop_tc$so, method = "lin")
-)
-print(AIC_parent_saemix_methods)

+f_parent_saemix_dfop_tc$so <-
+  saemix::llgq.saemix(f_parent_saemix_dfop_tc$so)
+AIC_parent_saemix_methods <- c(
+  is = AIC(f_parent_saemix_dfop_tc$so, method = "is"),
+  gq = AIC(f_parent_saemix_dfop_tc$so, method = "gq"),
+  lin = AIC(f_parent_saemix_dfop_tc$so, method = "lin")
+)
+print(AIC_parent_saemix_methods)

    is     gq    lin 
 665.65 665.68 665.11

The AIC values based on importance sampling and Gaussian quadrature are very similar. Using linearisation is known to be less accurate, but still gives a similar value. In order to illustrate that the comparison of the three method depends on the degree of convergence obtained in the fit, the same comparison is shown below for the fit using the defaults for the number of iterations and the number of MCMC chains.

-f_parent_saemix_dfop_tc_defaults <- mkin::saem(f_parent_mkin_tc["DFOP", ])
-f_parent_saemix_dfop_tc_defaults$so <-
-  saemix::llgq.saemix(f_parent_saemix_dfop_tc_defaults$so)
-AIC_parent_saemix_methods_defaults <- c(
-  is = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "is"),
-  gq = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "gq"),
-  lin = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "lin")
-)
-print(AIC_parent_saemix_methods_defaults)

The AIC values based on importance sampling and Gaussian quadrature are very similar. Using linearisation is known to be less accurate, but still gives a similar value.

In order to illustrate that the comparison of the three method depends on the degree of convergence obtained in the fit, the same comparison is shown below for the fit using the defaults for the number of iterations and the number of MCMC chains.

When using OpenBlas for linear algebra, there is a large difference in the values obtained with Gaussian quadrature, so the larger number of iterations makes a lot of difference. When using the LAPACK version coming with Debian Bullseye, the AIC based on Gaussian quadrature is almost the same as the one obtained with the other methods, also when using defaults for the fit.

+f_parent_saemix_dfop_tc_defaults <- mkin::saem(f_parent_mkin_tc["DFOP", ])
+f_parent_saemix_dfop_tc_defaults$so <-
+  saemix::llgq.saemix(f_parent_saemix_dfop_tc_defaults$so)
+AIC_parent_saemix_methods_defaults <- c(
+  is = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "is"),
+  gq = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "gq"),
+  lin = AIC(f_parent_saemix_dfop_tc_defaults$so, method = "lin")
+)
+print(AIC_parent_saemix_methods_defaults)

    is     gq    lin 
 668.27 718.36 666.49

nlmixr -

In the last years, a lot of effort has been put into the nlmixr package which is designed for pharmacokinetics, where nonlinear mixed-effects models are routinely used, but which can also be used for related data like chemical degradation data. A current development branch of the mkin package provides an interface between mkin and nlmixr. Here, we check if we get equivalent results when using a refined version of the First Order Conditional Estimation (FOCE) algorithm used in nlme, namely the First Order Conditional Estimation with Interaction (FOCEI), and the SAEM algorithm as implemented in nlmixr.

First, the focei algorithm is used for the four model combinations.

-library(nlmixr)
-f_parent_nlmixr_focei_sfo_const <- nlmixr(f_parent_mkin_const["SFO", ], est = "focei")
-f_parent_nlmixr_focei_sfo_tc <- nlmixr(f_parent_mkin_tc["SFO", ], est = "focei")
-f_parent_nlmixr_focei_dfop_const <- nlmixr(f_parent_mkin_const["DFOP", ], est = "focei")
-f_parent_nlmixr_focei_dfop_tc<- nlmixr(f_parent_mkin_tc["DFOP", ], est = "focei")

For the SFO model with constant variance, the AIC values are the same, for the DFOP model, there are significant differences between the AIC values. These may be caused by different solutions that are found, but also by the fact that the AIC values for the nlmixr fits are calculated based on Gaussian quadrature, not on linearisation.

-aic_nlmixr_focei <- sapply(
-  list(f_parent_nlmixr_focei_sfo_const$nm, f_parent_nlmixr_focei_sfo_tc$nm,
-    f_parent_nlmixr_focei_dfop_const$nm, f_parent_nlmixr_focei_dfop_tc$nm),
-  AIC)
-aic_nlme <- sapply(
-  list(f_parent_nlme_sfo_const, NA, f_parent_nlme_sfo_tc, f_parent_nlme_dfop_tc),
-  function(x) if (is.na(x[1])) NA else AIC(x))
-aic_nlme_nlmixr_focei <- data.frame(
-  "Degradation model" = c("SFO", "SFO", "DFOP", "DFOP"),
-  "Error model" = rep(c("constant variance", "two-component"), 2),
-  "AIC (nlme)" = aic_nlme,
-  "AIC (nlmixr with FOCEI)" = aic_nlmixr_focei,
-  check.names = FALSE
-)
-print(aic_nlme_nlmixr_focei)

  Degradation model       Error model AIC (nlme) AIC (nlmixr with FOCEI)
-1               SFO constant variance     796.60                  796.60
-2               SFO     two-component         NA                  798.64
-3              DFOP constant variance     798.60                  745.87
-4              DFOP     two-component     671.91                  740.42

Secondly, we use the SAEM estimation routine and check the convergence plots. The control parameters, which were also used for the saemix fits, are defined beforehand.

-nlmixr_saem_control_800 <- saemControl(logLik = TRUE,
-  nBurn = 800, nEm = 300, nmc = 15)
-nlmixr_saem_control_moreiter <- saemControl(logLik = TRUE,
-  nBurn = 1600, nEm = 300, nmc = 15)
-nlmixr_saem_control_10k <- saemControl(logLik = TRUE,
-  nBurn = 10000, nEm = 1000, nmc = 15)

Then we fit SFO with constant variance

-f_parent_nlmixr_saem_sfo_const <- nlmixr(f_parent_mkin_const["SFO", ], est = "saem",
-  control = nlmixr_saem_control_800)
-traceplot(f_parent_nlmixr_saem_sfo_const$nm)

and SFO with two-component error.

-f_parent_nlmixr_saem_sfo_tc <- nlmixr(f_parent_mkin_tc["SFO", ], est = "saem",
-  control = nlmixr_saem_control_800)
-traceplot(f_parent_nlmixr_saem_sfo_tc$nm)

For DFOP with constant variance, the convergence plots show considerable instability of the fit, which indicates overparameterisation which was already observed above for this model combination. Also note that the variance of k2 approximates zero, which was already observed in the saemix fits of the DFOP model.

-f_parent_nlmixr_saem_dfop_const <- nlmixr(f_parent_mkin_const["DFOP", ], est = "saem",
-  control = nlmixr_saem_control_800)
-traceplot(f_parent_nlmixr_saem_dfop_const$nm)

For DFOP with two-component error, a less erratic convergence is seen, but the variance of k2 again approximates zero.

-f_parent_nlmixr_saem_dfop_tc <- nlmixr(f_parent_mkin_tc["DFOP", ], est = "saem",
-  control = nlmixr_saem_control_800)
-traceplot(f_parent_nlmixr_saem_dfop_tc$nm)

To check if an increase in the number of iterations improves the fit, we repeat the fit with 1000 iterations for the burn in phase and 300 iterations for the second phase.

-f_parent_nlmixr_saem_dfop_tc_moreiter <- nlmixr(f_parent_mkin_tc["DFOP", ], est = "saem",
-  control = nlmixr_saem_control_moreiter)
-traceplot(f_parent_nlmixr_saem_dfop_tc_moreiter$nm)

Here the fit looks very similar, but we will see below that it shows a higher AIC than the fit with 800 iterations in the burn in phase. Next we choose 10 000 iterations for the burn in phase and 1000 iterations for the second phase for comparison with saemix.

-f_parent_nlmixr_saem_dfop_tc_10k <- nlmixr(f_parent_mkin_tc["DFOP", ], est = "saem",
-  control = nlmixr_saem_control_10k)
-traceplot(f_parent_nlmixr_saem_dfop_tc_10k$nm)

The AIC values are internally calculated using Gaussian quadrature.

-AIC(f_parent_nlmixr_saem_sfo_const$nm, f_parent_nlmixr_saem_sfo_tc$nm,
-  f_parent_nlmixr_saem_dfop_const$nm, f_parent_nlmixr_saem_dfop_tc$nm,
-  f_parent_nlmixr_saem_dfop_tc_moreiter$nm,
-  f_parent_nlmixr_saem_dfop_tc_10k$nm)

                                         df     AIC
-f_parent_nlmixr_saem_sfo_const$nm         5  798.71
-f_parent_nlmixr_saem_sfo_tc$nm            6  808.64
-f_parent_nlmixr_saem_dfop_const$nm        9 1995.96
-f_parent_nlmixr_saem_dfop_tc$nm          10  664.96
-f_parent_nlmixr_saem_dfop_tc_moreiter$nm 10 4464.93
-f_parent_nlmixr_saem_dfop_tc_10k$nm      10     Inf

We can see that again, the DFOP/tc model shows the best goodness of fit. However, increasing the number of burn-in iterations from 800 to 1600 results in a higher AIC. If we further increase the number of iterations to 10 000 (burn-in) and 1000 (second phase), the AIC cannot be calculated for the nlmixr/saem fit, confirming that this fit does not converge properly with the SAEM algorithm.

Comparison -

The following table gives the AIC values obtained with the three packages using the same control parameters (800 iterations burn-in, 300 iterations second phase, 15 chains).

-AIC_all <- data.frame(
-  check.names = FALSE,
-  "Degradation model" = c("SFO", "SFO", "DFOP", "DFOP"),
-  "Error model" = c("const", "tc", "const", "tc"),
-  nlme = c(AIC(f_parent_nlme_sfo_const), AIC(f_parent_nlme_sfo_tc), NA, AIC(f_parent_nlme_dfop_tc)),
-  nlmixr_focei = sapply(list(f_parent_nlmixr_focei_sfo_const$nm, f_parent_nlmixr_focei_sfo_tc$nm,
-  f_parent_nlmixr_focei_dfop_const$nm, f_parent_nlmixr_focei_dfop_tc$nm), AIC),
-  saemix = sapply(list(f_parent_saemix_sfo_const$so, f_parent_saemix_sfo_tc$so,
-    f_parent_saemix_dfop_const$so, f_parent_saemix_dfop_tc$so), AIC),
-  nlmixr_saem = sapply(list(f_parent_nlmixr_saem_sfo_const$nm, f_parent_nlmixr_saem_sfo_tc$nm,
-  f_parent_nlmixr_saem_dfop_const$nm, f_parent_nlmixr_saem_dfop_tc$nm), AIC)
-)
-kable(AIC_all)

Comparison +

The following table gives the AIC values obtained with both backend packages using the same control parameters (800 iterations burn-in, 300 iterations second phase, 15 chains).

+AIC_all <- data.frame(
+  check.names = FALSE,
+  "Degradation model" = c("SFO", "SFO", "DFOP", "DFOP"),
+  "Error model" = c("const", "tc", "const", "tc"),
+  nlme = c(AIC(f_parent_nlme_sfo_const), AIC(f_parent_nlme_sfo_tc), NA, AIC(f_parent_nlme_dfop_tc)),
+  saemix_lin = sapply(list(f_parent_saemix_sfo_const$so, f_parent_saemix_sfo_tc$so,
+    f_parent_saemix_dfop_const$so, f_parent_saemix_dfop_tc$so), AIC, method = "lin"),
+  saemix_is = sapply(list(f_parent_saemix_sfo_const$so, f_parent_saemix_sfo_tc$so,
+    f_parent_saemix_dfop_const$so, f_parent_saemix_dfop_tc$so), AIC, method = "is")
+)
+kable(AIC_all)

- - - + + @@ -435,36 +400,81 @@ f_parent_nlmixr_saem_dfop_tc_10k$nm 10 Inf - - + - - + - - + -

Degradation model	Error model	nlme	nlmixr_focei	saemix	nlmixr_saem	saemix_lin
796.60	796.60	796.38	798.71
SFO	tc	798.60	798.64	798.60	798.38	808.64
DFOP	const	NA	745.87	671.98	705.75	1995.96
DFOP	tc	671.91	740.42	665.11	665.65	664.96

Conclusion +

A more detailed analysis of the dimethenamid dataset confirmed that the DFOP model provides the most appropriate description of the decline of the parent compound in these data. On the other hand, closer inspection of the results revealed that the variability of the k2 parameter across the population of soils is ill-defined. This coincides with the observation that this parameter cannot robustly be quantified for some of the soils.

Regarding the regulatory use of these data, it is claimed that an improved characterisation of the mean parameter values across the population is obtained using the nonlinear mixed-effects models presented here. However, attempts to quantify the variability of the slower rate constant of the biphasic decline of dimethenamid indicate that the data are not sufficient to characterise this variability to a satisfactory precision.

Session Info +

+sessionInfo()

R version 4.2.1 (2022-06-23)
+Platform: x86_64-pc-linux-gnu (64-bit)
+Running under: Debian GNU/Linux 11 (bullseye)
+
+Matrix products: default
+BLAS:   /usr/lib/x86_64-linux-gnu/openblas-serial/libblas.so.3
+LAPACK: /usr/lib/x86_64-linux-gnu/openblas-serial/libopenblas-r0.3.13.so
+
+locale:
+ [1] LC_CTYPE=de_DE.UTF-8       LC_NUMERIC=C              
+ [3] LC_TIME=C                  LC_COLLATE=de_DE.UTF-8    
+ [5] LC_MONETARY=de_DE.UTF-8    LC_MESSAGES=de_DE.UTF-8   
+ [7] LC_PAPER=de_DE.UTF-8       LC_NAME=C                 
+ [9] LC_ADDRESS=C               LC_TELEPHONE=C            
+[11] LC_MEASUREMENT=de_DE.UTF-8 LC_IDENTIFICATION=C       
+
+attached base packages:
+[1] stats     graphics  grDevices utils     datasets  methods   base     
+
+other attached packages:
+[1] saemix_3.1   npde_3.2     nlme_3.1-158 mkin_1.1.2   knitr_1.39  
+
+loaded via a namespace (and not attached):
+ [1] deSolve_1.33      zoo_1.8-10        tidyselect_1.1.2  xfun_0.31        
+ [5] bslib_0.4.0       purrr_0.3.4       lattice_0.20-45   colorspace_2.0-3 
+ [9] vctrs_0.4.1       generics_0.1.3    htmltools_0.5.3   yaml_2.3.5       
+[13] utf8_1.2.2        rlang_1.0.4       pkgdown_2.0.6     jquerylib_0.1.4  
+[17] pillar_1.8.0      glue_1.6.2        DBI_1.1.3         lifecycle_1.0.1  
+[21] stringr_1.4.0     munsell_0.5.0     gtable_0.3.0      ragg_1.2.2       
+[25] codetools_0.2-18  memoise_2.0.1     evaluate_0.15     fastmap_1.1.0    
+[29] lmtest_0.9-40     parallel_4.2.1    fansi_1.0.3       highr_0.9        
+[33] scales_1.2.0      cachem_1.0.6      desc_1.4.1        jsonlite_1.8.0   
+[37] systemfonts_1.0.4 fs_1.5.2          textshaping_0.3.6 gridExtra_2.3    
+[41] ggplot2_3.3.6     digest_0.6.29     stringi_1.7.8     dplyr_1.0.9      
+[45] grid_4.2.1        rprojroot_2.0.3   cli_3.3.0         tools_4.2.1      
+[49] magrittr_2.0.3    sass_0.4.2        tibble_3.1.8      pkgconfig_2.0.3  
+[53] assertthat_0.2.1  rmarkdown_2.14.3  R6_2.5.1          mclust_5.4.10    
+[57] compiler_4.2.1

References @@ -501,7 +511,7 @@ f_parent_nlmixr_saem_dfop_tc_10k$nm 10 Inf

-
Site built with pkgdown 2.0.2.
+
Site built with pkgdown 2.0.6.

diff --git a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const-1.png b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const-1.png index 950abf27..4999e72c 100644 Binary files a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const-1.png and b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const-1.png differ diff --git a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const_test-1.png b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const_test-1.png index 66481bc1..b59764b1 100644 Binary files a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const_test-1.png and b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_const_test-1.png differ diff --git a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_tc_test-1.png b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_tc_test-1.png index 557ccc45..da7ceeb6 100644 Binary files a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_tc_test-1.png and b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_dfop_tc_test-1.png differ diff --git a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_sfo_const-1.png b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_sfo_const-1.png index 089ab64d..467c3c1a 100644 Binary files a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_sfo_const-1.png and b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/f_parent_mkin_sfo_const-1.png differ diff --git a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/plot_parent_nlme-1.png b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/plot_parent_nlme-1.png index 849296d9..4d2dc94e 100644 Binary files a/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/plot_parent_nlme-1.png and b/docs/dev/articles/web_only/dimethenamid_2018_files/figure-html/plot_parent_nlme-1.png differ -- cgit v1.2.1

Example evaluation of FOCUS Laboratory Data L1 to L3

Johannes Ranke

Johannes Ranke

Last change 17 November 2016 (rebuilt 2021-02-15)

Last change 18 May 2022 (rebuilt 2022-08-10)

-Laboratory Data L1