Calculate Akaike weights for model averaging

Arguments

Date: Wed, 13 Nov 2019 21:15:35 +0100 Subject: Calculate Akaike weights --- NAMESPACE | 3 + NEWS.md | 2 + R/aw.R | 60 +++++++++++ _pkgdown.yml | 1 + docs/news/index.html | 1 + docs/reference/aw.html | 214 ++++++++++++++++++++++++++++++++++++++++ docs/reference/index.html | 6 ++ docs/sitemap.xml | 3 + man/aw.Rd | 47 +++++++++ test.log | 21 ++-- tests/testthat/FOCUS_2006_D.csf | 2 +- tests/testthat/test_aw.R | 12 +++ 12 files changed, 361 insertions(+), 11 deletions(-) create mode 100644 R/aw.R create mode 100644 docs/reference/aw.html create mode 100644 man/aw.Rd create mode 100644 tests/testthat/test_aw.R (limited to 'docs/reference/index.html') diff --git a/NAMESPACE b/NAMESPACE index e561621b..26995055 100644 --- a/NAMESPACE +++ b/NAMESPACE @@ -3,6 +3,8 @@ S3method("[",mmkin) S3method(AIC,mmkin) S3method(BIC,mmkin) +S3method(aw,mkinfit) +S3method(aw,mmkin) S3method(confint,mkinfit) S3method(loftest,mkinfit) S3method(logLik,mkinfit) @@ -30,6 +32,7 @@ export(IORE.solution) export(SFO.solution) export(SFORB.solution) export(add_err) +export(aw) export(backtransform_odeparms) export(endpoints) export(ilr) diff --git a/NEWS.md b/NEWS.md index 965105f4..28cf76ad 100644 --- a/NEWS.md +++ b/NEWS.md @@ -1,5 +1,7 @@ # mkin 0.9.49.8 (unreleased) +- 'aw': Generic function for calculating Akaike weights, methods for mkinfit objects and mmkin columns + - 'loftest': Add a lack-of-fit test - 'plot_res', 'plot_sep' and 'mkinerrplot': Add the possibility to show standardized residuals and make it the default for fits with error models other than 'const' diff --git a/R/aw.R b/R/aw.R new file mode 100644 index 00000000..24078f38 --- /dev/null +++ b/R/aw.R @@ -0,0 +1,60 @@ +#' Calculate Akaike weights for model averaging +#' +#' Akaike weights are calculated based on the relative +#' expected Kullback-Leibler information as specified +#' by Burnham and Anderson (2004). +#' +#' @param object An mmkin column object, containing two or more +#' \code{\link{mkinfit}} models that have been fitted to the same data, +#' or an mkinfit object. In the latter case, further mkinfit +#' objects fitted to the same data should be specified +#' as dots arguments. +#' @param \dots Not used in the method for mmkin column objects, +#' further mkinfit objects in the method for mkinfit objects. +#' @references Burnham KP and Anderson DR (2004) Multimodel +#' Inference: Understanding AIC and BIC in Model Selection +#' Sociological Methods & Research 33(2) 261-304 +#' @examples +#' \dontrun{ +#' f_sfo <- mkinfit("SFO", FOCUS_2006_D, quiet = TRUE) +#' f_dfop <- mkinfit("DFOP", FOCUS_2006_D, quiet = TRUE) +#' aw_sfo_dfop <- aw(f_sfo, f_dfop) +#' sum(aw_sfo_dfop) +#' aw_sfo_dfop # SFO gets more weight as it has less parameters and a similar fit +#' f <- mmkin(c("SFO", "FOMC", "DFOP"), list("FOCUS D" = FOCUS_2006_D), cores = 1, quiet = TRUE) +#' aw(f) +#' sum(aw(f)) +#' aw(f[c("SFO", "DFOP")]) +#' } +#' @export +aw <- function(object, ...) UseMethod("aw") + +#' @export +#' @rdname aw +aw.mkinfit <- function(object, ...) { + oo <- list(...) + data_object <- object$data[c("time", "variable", "observed")] + for (i in seq_along(oo)) { + if (!inherits(oo[[i]], "mkinfit")) stop("Please supply only mkinfit objects") + data_other_object <- oo[[i]]$data[c("time", "variable", "observed")] + if (!identical(data_object, data_other_object)) { + stop("It seems that the mkinfit objects have not all been fitted to the same data") + } + } + all_objects <- list(object, ...) + AIC_all <- sapply(all_objects, AIC) + delta_i <- AIC_all - min(AIC_all) + denom <- sum(exp(-delta_i/2)) + w_i <- exp(-delta_i/2) / denom + return(w_i) +} + +#' @export +#' @rdname aw +aw.mmkin <- function(object, ...) { + if (ncol(object) > 1) stop("Please supply an mmkin column object") + do.call(aw, object) +} + + + diff --git a/_pkgdown.yml b/_pkgdown.yml index c298256f..6bb05b3e 100644 --- a/_pkgdown.yml +++ b/_pkgdown.yml @@ -25,6 +25,7 @@ reference: - loftest - mkinerrmin - endpoints + - aw - CAKE_export - title: Work with mmkin objects desc: Functions working with aggregated results diff --git a/docs/news/index.html b/docs/news/index.html index 9aa2e18b..6b0b89fa 100644 --- a/docs/news/index.html +++ b/docs/news/index.html @@ -134,6 +134,7 @@ mkin 0.9.49.8 (unreleased) Unreleased

‘aw’: Generic function for calculating Akaike weights, methods for mkinfit objects and mmkin columns
‘loftest’: Add a lack-of-fit test
‘plot_res’, ‘plot_sep’ and ‘mkinerrplot’: Add the possibility to show standardized residuals and make it the default for fits with error models other than ‘const’
‘lrtest.mkinfit’: Improve naming of the compared fits in the case of fixed parameters

Calculate Akaike weights for model averaging — aw • mkin

+ + + + +

+ +

+ + +

Akaike weights are calculated based on the relative +expected Kullback-Leibler information as specified +by Burnham and Anderson (2004).

+ +

aw(object, ...)
+
+# S3 method for mkinfit
+aw(object, ...)
+
+# S3 method for mmkin
+aw(object, ...)

+ +

Arguments

object	An `mkinfit` object	An `mkinfit` object, or an `mmkin` column +object containing two fits to the same data.
object_2

object

An mkinfit object

An mkinfit object, or an mmkin column +object containing two fits to the same data.

object_2

object	An mmkin column object, containing two or more +`mkinfit` models that have been fitted to the same data, +or an mkinfit object. In the latter case, further mkinfit +objects fitted to the same data should be specified +as dots arguments.
...	Not used in the method for mmkin column objects, +further mkinfit objects in the method for mkinfit objects.

object

An mmkin column object, containing two or more +mkinfit models that have been fitted to the same data, +or an mkinfit object. In the latter case, further mkinfit +objects fitted to the same data should be specified +as dots arguments.

...

Not used in the method for mmkin column objects, +further mkinfit objects in the method for mkinfit objects.

Examples

# \dontrun{ +f_sfo <- mkinfit("SFO", FOCUS_2006_D, quiet = TRUE) +f_dfop <- mkinfit("DFOP", FOCUS_2006_D, quiet = TRUE) +aw_sfo_dfop <- aw(f_sfo, f_dfop) +sum(aw_sfo_dfop)

#> [1] 1

aw_sfo_dfop # SFO gets more weight as it has less parameters and a similar fit

#> [1] 0.5970258 0.4029742

f <- mmkin(c("SFO", "FOMC", "DFOP"), list("FOCUS D" = FOCUS_2006_D), cores = 1, quiet = TRUE) +aw(f)

#> [1] 0.4808722 0.1945539 0.3245740

sum(aw(f))

#> [1] 1

aw(f[c("SFO", "DFOP")])

#> [1] 0.5970258 0.4029742

# } +

Calculate Akaike weights for model averaging

https://pkgdown.jrwb.de/mkin/reference/add_err.html

+ https://pkgdown.jrwb.de/mkin/reference/aw.html +

https://pkgdown.jrwb.de/mkin/reference/confint.mkinfit.html

Date: Mon, 9 Dec 2019 15:21:11 +0100 Subject: Consistently use "two-component error model" instead of "two component error model" --- R/sigma_twocomp.R | 2 +- docs/reference/index.html | 2 +- docs/reference/sigma_twocomp.html | 8 ++++---- man/sigma_twocomp.Rd | 2 +- 4 files changed, 7 insertions(+), 7 deletions(-) (limited to 'docs/reference/index.html') diff --git a/R/sigma_twocomp.R b/R/sigma_twocomp.R index c9a15aa8..1e012d15 100644 --- a/R/sigma_twocomp.R +++ b/R/sigma_twocomp.R @@ -1,4 +1,4 @@ -#' Two component error model +#' Two-component error model #' #' Function describing the standard deviation of the measurement error in #' dependence of the measured value \eqn{y}: diff --git a/docs/reference/index.html b/docs/reference/index.html index 3d417267..73dfbe4c 100644 --- a/docs/reference/index.html +++ b/docs/reference/index.html @@ -444,7 +444,7 @@ kinetic models fitted with mkinfit

sigma_twocomp()

Two component error model

Two-component error model

diff --git a/docs/reference/sigma_twocomp.html b/docs/reference/sigma_twocomp.html index d95e1c28..1795e6c3 100644 --- a/docs/reference/sigma_twocomp.html +++ b/docs/reference/sigma_twocomp.html @@ -6,7 +6,7 @@ -Two component error model — sigma_twocomp • mkin +Two-component error model — sigma_twocomp • mkin @@ -35,7 +35,7 @@ - + @@ -70,7 +70,7 @@ dependence of the measured value $y$:" /> @@ -128,7 +128,7 @@ dependence of the measured value $y$:" />

diff --git a/man/sigma_twocomp.Rd b/man/sigma_twocomp.Rd index 3e7854f1..0004144f 100644 --- a/man/sigma_twocomp.Rd +++ b/man/sigma_twocomp.Rd @@ -2,7 +2,7 @@ % Please edit documentation in R/sigma_twocomp.R \name{sigma_twocomp} \alias{sigma_twocomp} -\title{Two component error model} +\title{Two-component error model} \usage{ sigma_twocomp(y, sigma_low, rsd_high) } -- cgit v1.2.3 From 70fe6d14e27fa8fb0634856ecc45a27f4f689d88 Mon Sep 17 00:00:00 2001 From: Johannes Ranke Date: Thu, 9 Jan 2020 17:50:23 +0100 Subject: Remove unused FME package from DESCRIPTION text --- DESCRIPTION | 13 ++++++------- build.log | 2 +- docs/404.html | 2 +- docs/articles/index.html | 2 +- docs/authors.html | 2 +- docs/index.html | 11 +++++------ docs/news/index.html | 2 +- docs/reference/index.html | 2 +- docs/reference/parms.html | 2 +- 9 files changed, 18 insertions(+), 20 deletions(-) (limited to 'docs/reference/index.html') diff --git a/DESCRIPTION b/DESCRIPTION index f9bed994..0bc78ed2 100644 --- a/DESCRIPTION +++ b/DESCRIPTION @@ -1,8 +1,8 @@ Package: mkin Type: Package Title: Kinetic Evaluation of Chemical Degradation Data -Version: 0.9.49.8 -Date: 2020-01-08 +Version: 0.9.49.9 +Date: 2020-01-09 Authors@R: c(person("Johannes", "Ranke", role = c("aut", "cre", "cph"), email = "jranke@uni-bremen.de", comment = c(ORCID = "0000-0003-4371-6538")), @@ -12,11 +12,10 @@ Authors@R: c(person("Johannes", "Ranke", role = c("aut", "cre", "cph"), Description: Calculation routines based on the FOCUS Kinetics Report (2006, 2014). Includes a function for conveniently defining differential equation models, model solution based on eigenvalues if possible or using numerical - solvers and a choice of the optimisation methods made available by the 'FME' - package. If a C compiler (on windows: 'Rtools') is installed, differential - equation models are solved using compiled C functions. Please note that no - warranty is implied for correctness of results or fitness for a particular - purpose. + solvers. If a C compiler (on windows: 'Rtools') is installed, differential + equation models are solved using automatically generated C functions. Please + note that no warranty is implied for correctness of results or fitness for a + particular purpose. Imports: stats, graphics, methods, deSolve, R6, inline, parallel, numDeriv, lmtest Suggests: knitr, rbenchmark, tikzDevice, testthat, rmarkdown, covr, vdiffr, diff --git a/build.log b/build.log index ffd56a90..b0765ebb 100644 --- a/build.log +++ b/build.log @@ -6,5 +6,5 @@ * checking for LF line-endings in source and make files and shell scripts * checking for empty or unneeded directories * looking to see if a ‘data/datalist’ file should be added -* building ‘mkin_0.9.49.8.tar.gz’ +* building ‘mkin_0.9.49.9.tar.gz’ diff --git a/docs/404.html b/docs/404.html index 2a71e496..3658a137 100644 --- a/docs/404.html +++ b/docs/404.html @@ -67,7 +67,7 @@

diff --git a/docs/articles/index.html b/docs/articles/index.html index ac5f5df1..3e181a2c 100644 --- a/docs/articles/index.html +++ b/docs/articles/index.html @@ -67,7 +67,7 @@

diff --git a/docs/authors.html b/docs/authors.html index fb50f268..2bb2557a 100644 --- a/docs/authors.html +++ b/docs/authors.html @@ -67,7 +67,7 @@ diff --git a/docs/index.html b/docs/index.html index fcb9c90b..fe334eb8 100644 --- a/docs/index.html +++ b/docs/index.html @@ -14,11 +14,10 @@ + solvers. If a C compiler (on windows: Rtools) is installed, differential + equation models are solved using automatically generated C functions. Please + note that no warranty is implied for correctness of results or fitness for a + particular purpose.">

Arguments

Calculate Akaike weights for model averaging

Arguments

References

Examples

Contents

Two component error model

Two-component error model