This function sets up a nonlinear mixed effects model for an mmkin row object for use with the saemix package. An mmkin row object is essentially a list of mkinfit objects that have been obtained by fitting the same model to a list of datasets.
saemix_model(object, cores = parallel::detectCores()) saemix_data(object, ...)
Arguments
| object | An mmkin row object containing several fits of the same model to different datasets |
|---|---|
| cores | The number of cores to be used for multicore processing. On Windows machines, cores > 1 is currently not supported. |
| ... | Further parameters passed to saemix::saemixData |
Value
An saemix::SaemixModel object.
An saemix::SaemixData object.
Details
Starting values for the fixed effects (population mean parameters, argument psi0 of
saemix::saemixModel() are the mean values of the parameters found using
mmkin. Starting variances of the random effects (argument omega.init) are the
variances of the deviations of the parameters from these mean values.
Examples
ds <- lapply(experimental_data_for_UBA_2019[6:10], function(x) subset(x$data[c("name", "time", "value")])) names(ds) <- paste("Dataset", 6:10) sfo_sfo <- mkinmod(parent = mkinsub("SFO", "A1"), A1 = mkinsub("SFO"))#>#> #>m_saemix <- saemix_model(f_mmkin, cores = 1)#> #> #> The following SaemixModel object was successfully created: #> #> Nonlinear mixed-effects model #> Model function: Mixed model generated from mmkin object Model type: structural #> function (psi, id, xidep) #> { #> uid <- unique(id) #> res_list <- parallel::mclapply(uid, function(i) { #> transparms_optim <- psi[i, ] #> names(transparms_optim) <- names(degparms_optim) #> odeini_optim <- transparms_optim[odeini_optim_parm_names] #> names(odeini_optim) <- gsub("_0$", "", odeini_optim_parm_names) #> odeini <- c(odeini_optim, odeini_fixed)[names(mkin_model$diffs)] #> ode_transparms_optim_names <- setdiff(names(transparms_optim), #> odeini_optim_parm_names) #> odeparms_optim <- backtransform_odeparms(transparms_optim[ode_transparms_optim_names], #> mkin_model, transform_rates = object[[1]]$transform_rates, #> transform_fractions = object[[1]]$transform_fractions) #> odeparms <- c(odeparms_optim, odeparms_fixed) #> xidep_i <- subset(xidep, id == i) #> if (analytical) { #> out_values <- mkin_model$deg_func(xidep_i, odeini, #> odeparms) #> } #> else { #> i_time <- xidep_i$time #> i_name <- xidep_i$name #> out_wide <- mkinpredict(mkin_model, odeparms = odeparms, #> odeini = odeini, solution_type = object[[1]]$solution_type, #> outtimes = sort(unique(i_time))) #> out_index <- cbind(as.character(i_time), as.character(i_name)) #> out_values <- out_wide[out_index] #> } #> return(out_values) #> }, mc.cores = cores) #> res <- unlist(res_list) #> return(res) #> } #> <bytecode: 0x555555c89340> #> <environment: 0x555555c82680> #> Nb of parameters: 4 #> parameter names: parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> distribution: #> Parameter Distribution Estimated #> [1,] parent_0 normal Estimated #> [2,] log_k_parent normal Estimated #> [3,] log_k_A1 normal Estimated #> [4,] f_parent_ilr_1 normal Estimated #> Variance-covariance matrix: #> parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> parent_0 1 0 0 0 #> log_k_parent 0 1 0 0 #> log_k_A1 0 0 1 0 #> f_parent_ilr_1 0 0 0 1 #> Error model: constant , initial values: a.1=4.97259024646577 #> No covariate in the model. #> Initial values #> parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> Pop.CondInit 86.53449 -3.207005 -3.060308 -1.920449d_saemix <- saemix_data(f_mmkin)#> #> #> The following SaemixData object was successfully created: #> #> Object of class SaemixData #> longitudinal data for use with the SAEM algorithm #> Dataset ds_saemix #> Structured data: value ~ time + name | ds #> X variable for graphs: time ()saemix_options <- list(seed = 123456, save = FALSE, save.graphs = FALSE, displayProgress = FALSE, nbiter.saemix = c(200, 80)) f_saemix <- saemix(m_saemix, d_saemix, saemix_options)#> Running main SAEM algorithm #> [1] "Thu Oct 15 14:51:26 2020" #> .. #> Minimisation finished #> [1] "Thu Oct 15 14:53:18 2020"#> Nonlinear mixed-effects model fit by the SAEM algorithm #> ----------------------------------- #> ---- Data ---- #> ----------------------------------- #> Object of class SaemixData #> longitudinal data for use with the SAEM algorithm #> Dataset ds_saemix #> Structured data: value ~ time + name | ds #> X variable for graphs: time () #> Dataset characteristics: #> number of subjects: 5 #> number of observations: 170 #> average/min/max nb obs: 34.00 / 30 / 38 #> First 10 lines of data: #> ds time name value mdv cens occ ytype #> 1 Dataset 6 0 parent 97.2 0 0 1 1 #> 2 Dataset 6 0 parent 96.4 0 0 1 1 #> 3 Dataset 6 3 parent 71.1 0 0 1 1 #> 4 Dataset 6 3 parent 69.2 0 0 1 1 #> 5 Dataset 6 6 parent 58.1 0 0 1 1 #> 6 Dataset 6 6 parent 56.6 0 0 1 1 #> 7 Dataset 6 10 parent 44.4 0 0 1 1 #> 8 Dataset 6 10 parent 43.4 0 0 1 1 #> 9 Dataset 6 20 parent 33.3 0 0 1 1 #> 10 Dataset 6 20 parent 29.2 0 0 1 1 #> ----------------------------------- #> ---- Model ---- #> ----------------------------------- #> Nonlinear mixed-effects model #> Model function: Mixed model generated from mmkin object Model type: structural #> function (psi, id, xidep) #> { #> uid <- unique(id) #> res_list <- parallel::mclapply(uid, function(i) { #> transparms_optim <- psi[i, ] #> names(transparms_optim) <- names(degparms_optim) #> odeini_optim <- transparms_optim[odeini_optim_parm_names] #> names(odeini_optim) <- gsub("_0$", "", odeini_optim_parm_names) #> odeini <- c(odeini_optim, odeini_fixed)[names(mkin_model$diffs)] #> ode_transparms_optim_names <- setdiff(names(transparms_optim), #> odeini_optim_parm_names) #> odeparms_optim <- backtransform_odeparms(transparms_optim[ode_transparms_optim_names], #> mkin_model, transform_rates = object[[1]]$transform_rates, #> transform_fractions = object[[1]]$transform_fractions) #> odeparms <- c(odeparms_optim, odeparms_fixed) #> xidep_i <- subset(xidep, id == i) #> if (analytical) { #> out_values <- mkin_model$deg_func(xidep_i, odeini, #> odeparms) #> } #> else { #> i_time <- xidep_i$time #> i_name <- xidep_i$name #> out_wide <- mkinpredict(mkin_model, odeparms = odeparms, #> odeini = odeini, solution_type = object[[1]]$solution_type, #> outtimes = sort(unique(i_time))) #> out_index <- cbind(as.character(i_time), as.character(i_name)) #> out_values <- out_wide[out_index] #> } #> return(out_values) #> }, mc.cores = cores) #> res <- unlist(res_list) #> return(res) #> } #> <bytecode: 0x555555c89340> #> <environment: 0x555555c82680> #> Nb of parameters: 4 #> parameter names: parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> distribution: #> Parameter Distribution Estimated #> [1,] parent_0 normal Estimated #> [2,] log_k_parent normal Estimated #> [3,] log_k_A1 normal Estimated #> [4,] f_parent_ilr_1 normal Estimated #> Variance-covariance matrix: #> parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> parent_0 1 0 0 0 #> log_k_parent 0 1 0 0 #> log_k_A1 0 0 1 0 #> f_parent_ilr_1 0 0 0 1 #> Error model: constant , initial values: a.1=4.97259024646577 #> No covariate in the model. #> Initial values #> parent_0 log_k_parent log_k_A1 f_parent_ilr_1 #> Pop.CondInit 86.53449 -3.207005 -3.060308 -1.920449 #> ----------------------------------- #> ---- Key algorithm options ---- #> ----------------------------------- #> Estimation of individual parameters (MAP) #> Estimation of standard errors and linearised log-likelihood #> Estimation of log-likelihood by importance sampling #> Number of iterations: K1=200, K2=80 #> Number of chains: 10 #> Seed: 123456 #> Number of MCMC iterations for IS: 5000 #> Simulations: #> nb of simulated datasets used for npde: 1000 #> nb of simulated datasets used for VPC: 100 #> Input/output #> save the results to a file: FALSE #> save the graphs to files: FALSE #> ---------------------------------------------------- #> ---- Results ---- #> ---------------------------------------------------- #> ----------------- Fixed effects ------------------ #> ---------------------------------------------------- #> Parameter Estimate SE CV(%) #> parent_0 86.09 1.57 1.8 #> log_k_parent -3.21 0.59 18.5 #> log_k_A1 -4.69 0.31 6.6 #> f_parent_ilr_1 -0.34 0.30 89.2 #> a a.1 4.69 0.27 5.8 #> ---------------------------------------------------- #> ----------- Variance of random effects ----------- #> ---------------------------------------------------- #> Parameter Estimate SE CV(%) #> parent_0 omega2.parent_0 7.07 7.72 109 #> log_k_parent omega2.log_k_parent 1.75 1.11 63 #> log_k_A1 omega2.log_k_A1 0.28 0.28 99 #> f_parent_ilr_1 omega2.f_parent_ilr_1 0.39 0.27 71 #> ---------------------------------------------------- #> ------ Correlation matrix of random effects ------ #> ---------------------------------------------------- #> omega2.parent_0 omega2.log_k_parent omega2.log_k_A1 #> omega2.parent_0 1 0 0 #> omega2.log_k_parent 0 1 0 #> omega2.log_k_A1 0 0 1 #> omega2.f_parent_ilr_1 0 0 0 #> omega2.f_parent_ilr_1 #> omega2.parent_0 0 #> omega2.log_k_parent 0 #> omega2.log_k_A1 0 #> omega2.f_parent_ilr_1 1 #> ---------------------------------------------------- #> --------------- Statistical criteria ------------- #> ---------------------------------------------------- #> Likelihood computed by linearisation #> -2LL= 1064.35 #> AIC = 1082.35 #> BIC = 1078.835 #> #> Likelihood computed by importance sampling #> -2LL= 1063.475 #> AIC = 1081.475 #> BIC = 1077.96 #> ----------------------------------------------------#> Plotting convergence plots# } # Synthetic data with two-component error sampling_times = c(0, 1, 3, 7, 14, 28, 60, 90, 120) dt50_sfo_in <- c(80, 90, 100, 111.111, 125) k_in <- log(2) / dt50_sfo_in SFO <- mkinmod(parent = mkinsub("SFO")) pred_sfo <- function(k) { mkinpredict(SFO, c(k_parent = k), c(parent = 100), sampling_times) } ds_sfo_mean <- lapply(k_in, pred_sfo) set.seed(123456L) ds_sfo_syn <- lapply(ds_sfo_mean, function(ds) { add_err(ds, sdfunc = function(value) sqrt(1^2 + value^2 * 0.07^2), n = 1)[[1]] }) # \dontrun{ f_mmkin_syn <- mmkin("SFO", ds_sfo_syn, error_model = "tc", quiet = TRUE) # plot(f_mmkin_syn) m_saemix_tc <- saemix_model(f_mmkin_syn, cores = 1)#> #> #> The following SaemixModel object was successfully created: #> #> Nonlinear mixed-effects model #> Model function: Mixed model generated from mmkin object Model type: structural #> function (psi, id, xidep) #> { #> uid <- unique(id) #> res_list <- parallel::mclapply(uid, function(i) { #> transparms_optim <- psi[i, ] #> names(transparms_optim) <- names(degparms_optim) #> odeini_optim <- transparms_optim[odeini_optim_parm_names] #> names(odeini_optim) <- gsub("_0$", "", odeini_optim_parm_names) #> odeini <- c(odeini_optim, odeini_fixed)[names(mkin_model$diffs)] #> ode_transparms_optim_names <- setdiff(names(transparms_optim), #> odeini_optim_parm_names) #> odeparms_optim <- backtransform_odeparms(transparms_optim[ode_transparms_optim_names], #> mkin_model, transform_rates = object[[1]]$transform_rates, #> transform_fractions = object[[1]]$transform_fractions) #> odeparms <- c(odeparms_optim, odeparms_fixed) #> xidep_i <- subset(xidep, id == i) #> if (analytical) { #> out_values <- mkin_model$deg_func(xidep_i, odeini, #> odeparms) #> } #> else { #> i_time <- xidep_i$time #> i_name <- xidep_i$name #> out_wide <- mkinpredict(mkin_model, odeparms = odeparms, #> odeini = odeini, solution_type = object[[1]]$solution_type, #> outtimes = sort(unique(i_time))) #> out_index <- cbind(as.character(i_time), as.character(i_name)) #> out_values <- out_wide[out_index] #> } #> return(out_values) #> }, mc.cores = cores) #> res <- unlist(res_list) #> return(res) #> } #> <bytecode: 0x555555c89340> #> <environment: 0x55555df58cf8> #> Nb of parameters: 2 #> parameter names: parent_0 log_k_parent #> distribution: #> Parameter Distribution Estimated #> [1,] parent_0 normal Estimated #> [2,] log_k_parent normal Estimated #> Variance-covariance matrix: #> parent_0 log_k_parent #> parent_0 1 0 #> log_k_parent 0 1 #> Error model: combined , initial values: a.1=1.05209877924905 b.1=0.0586479225303944 #> No covariate in the model. #> Initial values #> parent_0 log_k_parent #> Pop.CondInit 100.315 -4.962075d_saemix_tc <- saemix_data(f_mmkin_syn)#> #> #> The following SaemixData object was successfully created: #> #> Object of class SaemixData #> longitudinal data for use with the SAEM algorithm #> Dataset ds_saemix #> Structured data: value ~ time + name | ds #> X variable for graphs: time ()#> Running main SAEM algorithm #> [1] "Thu Oct 15 14:53:35 2020" #> .. #> Minimisation finished #> [1] "Thu Oct 15 14:54:25 2020"#> Nonlinear mixed-effects model fit by the SAEM algorithm #> ----------------------------------- #> ---- Data ---- #> ----------------------------------- #> Object of class SaemixData #> longitudinal data for use with the SAEM algorithm #> Dataset ds_saemix #> Structured data: value ~ time + name | ds #> X variable for graphs: time () #> Dataset characteristics: #> number of subjects: 5 #> number of observations: 90 #> average/min/max nb obs: 18.00 / 18 / 18 #> First 10 lines of data: #> ds time name value mdv cens occ ytype #> 1 1 0 parent 105.9 0 0 1 1 #> 2 1 0 parent 98.0 0 0 1 1 #> 3 1 1 parent 96.6 0 0 1 1 #> 4 1 1 parent 99.8 0 0 1 1 #> 5 1 3 parent 113.0 0 0 1 1 #> 6 1 3 parent 103.2 0 0 1 1 #> 7 1 7 parent 102.9 0 0 1 1 #> 8 1 7 parent 110.8 0 0 1 1 #> 9 1 14 parent 95.9 0 0 1 1 #> 10 1 14 parent 85.9 0 0 1 1 #> ----------------------------------- #> ---- Model ---- #> ----------------------------------- #> Nonlinear mixed-effects model #> Model function: Mixed model generated from mmkin object Model type: structural #> function (psi, id, xidep) #> { #> uid <- unique(id) #> res_list <- parallel::mclapply(uid, function(i) { #> transparms_optim <- psi[i, ] #> names(transparms_optim) <- names(degparms_optim) #> odeini_optim <- transparms_optim[odeini_optim_parm_names] #> names(odeini_optim) <- gsub("_0$", "", odeini_optim_parm_names) #> odeini <- c(odeini_optim, odeini_fixed)[names(mkin_model$diffs)] #> ode_transparms_optim_names <- setdiff(names(transparms_optim), #> odeini_optim_parm_names) #> odeparms_optim <- backtransform_odeparms(transparms_optim[ode_transparms_optim_names], #> mkin_model, transform_rates = object[[1]]$transform_rates, #> transform_fractions = object[[1]]$transform_fractions) #> odeparms <- c(odeparms_optim, odeparms_fixed) #> xidep_i <- subset(xidep, id == i) #> if (analytical) { #> out_values <- mkin_model$deg_func(xidep_i, odeini, #> odeparms) #> } #> else { #> i_time <- xidep_i$time #> i_name <- xidep_i$name #> out_wide <- mkinpredict(mkin_model, odeparms = odeparms, #> odeini = odeini, solution_type = object[[1]]$solution_type, #> outtimes = sort(unique(i_time))) #> out_index <- cbind(as.character(i_time), as.character(i_name)) #> out_values <- out_wide[out_index] #> } #> return(out_values) #> }, mc.cores = cores) #> res <- unlist(res_list) #> return(res) #> } #> <bytecode: 0x555555c89340> #> <environment: 0x55555df58cf8> #> Nb of parameters: 2 #> parameter names: parent_0 log_k_parent #> distribution: #> Parameter Distribution Estimated #> [1,] parent_0 normal Estimated #> [2,] log_k_parent normal Estimated #> Variance-covariance matrix: #> parent_0 log_k_parent #> parent_0 1 0 #> log_k_parent 0 1 #> Error model: combined , initial values: a.1=1.05209877924905 b.1=0.0586479225303944 #> No covariate in the model. #> Initial values #> parent_0 log_k_parent #> Pop.CondInit 100.315 -4.962075 #> ----------------------------------- #> ---- Key algorithm options ---- #> ----------------------------------- #> Estimation of individual parameters (MAP) #> Estimation of standard errors and linearised log-likelihood #> Estimation of log-likelihood by importance sampling #> Number of iterations: K1=200, K2=80 #> Number of chains: 10 #> Seed: 123456 #> Number of MCMC iterations for IS: 5000 #> Simulations: #> nb of simulated datasets used for npde: 1000 #> nb of simulated datasets used for VPC: 100 #> Input/output #> save the results to a file: FALSE #> save the graphs to files: FALSE #> ---------------------------------------------------- #> ---- Results ---- #> ---------------------------------------------------- #> ----------------- Fixed effects ------------------ #> ---------------------------------------------------- #> Parameter Estimate SE CV(%) #> parent_0 100.232 1.266 1.3 #> log_k_parent -4.961 0.089 1.8 #> a a.1 -0.106 1.211 1142.0 #> b b.1 0.071 0.017 24.2 #> ---------------------------------------------------- #> ----------- Variance of random effects ----------- #> ---------------------------------------------------- #> Parameter Estimate SE CV(%) #> parent_0 omega2.parent_0 3.334 5.024 151 #> log_k_parent omega2.log_k_parent 0.036 0.024 68 #> ---------------------------------------------------- #> ------ Correlation matrix of random effects ------ #> ---------------------------------------------------- #> omega2.parent_0 omega2.log_k_parent #> omega2.parent_0 1 0 #> omega2.log_k_parent 0 1 #> ---------------------------------------------------- #> --------------- Statistical criteria ------------- #> ---------------------------------------------------- #> Likelihood computed by linearisation #> -2LL= 575.5586 #> AIC = 587.5586 #> BIC = 585.2153 #> #> Likelihood computed by importance sampling #> -2LL= 575.7797 #> AIC = 587.7797 #> BIC = 585.4364 #> ----------------------------------------------------#> Plotting convergence plots# }
