saem: Fit nonlinear mixed models with SAEM

Description

This function uses saemix::saemix() as a backend for fitting nonlinear mixed effects models created from mmkin row objects using the Stochastic Approximation Expectation Maximisation algorithm (SAEM).

Usage

saem(object, ...)
# S3 method for mmkin
saem(
  object,
  transformations = c("mkin", "saemix"),
  error_model = "auto",
  degparms_start = numeric(),
  test_log_parms = TRUE,
  conf.level = 0.6,
  solution_type = "auto",
  covariance.model = "auto",
  omega.init = "auto",
  covariates = NULL,
  covariate_models = NULL,
  no_random_effect = NULL,
  error.init = c(1, 1),
  nbiter.saemix = c(300, 100),
  control = list(displayProgress = FALSE, print = FALSE, nbiter.saemix = nbiter.saemix,
    save = FALSE, save.graphs = FALSE),
  verbose = FALSE,
  quiet = FALSE,
  ...
)
# S3 method for saem.mmkin
print(x, digits = max(3, getOption("digits") - 3), ...)
saemix_model(
  object,
  solution_type = "auto",
  transformations = c("mkin", "saemix"),
  error_model = "auto",
  degparms_start = numeric(),
  covariance.model = "auto",
  no_random_effect = NULL,
  omega.init = "auto",
  covariates = NULL,
  covariate_models = NULL,
  error.init = numeric(),
  test_log_parms = FALSE,
  conf.level = 0.6,
  verbose = FALSE,
  ...
)
saemix_data(object, covariates = NULL, verbose = FALSE, ...)

Value

An S3 object of class 'saem.mmkin', containing the fitted saemix::SaemixObject as a list component named 'so'. The object also inherits from 'mixed.mmkin'.

An saemix::SaemixModel object.

An saemix::SaemixData object.

Arguments

object: An mmkin row object containing several fits of the same mkinmod model to different datasets
...: Further parameters passed to saemix::saemixModel.
transformations: Per default, all parameter transformations are done in mkin. If this argument is set to 'saemix', parameter transformations are done in 'saemix' for the supported cases, i.e. (as of version 1.1.2) SFO, FOMC, DFOP and HS without fixing parent_0, and SFO or DFOP with one SFO metabolite.
error_model: Possibility to override the error model used in the mmkin object
degparms_start: Parameter values given as a named numeric vector will be used to override the starting values obtained from the 'mmkin' object.
test_log_parms: If TRUE, an attempt is made to use more robust starting values for population parameters fitted as log parameters in mkin (like rate constants) by only considering rate constants that pass the t-test when calculating mean degradation parameters using mean_degparms.
conf.level: Possibility to adjust the required confidence level for parameter that are tested if requested by 'test_log_parms'.
solution_type: Possibility to specify the solution type in case the automatic choice is not desired
covariance.model: Will be passed to saemix::saemixModel(). Per default, uncorrelated random effects are specified for all degradation parameters.
omega.init: Will be passed to saemix::saemixModel(). If using mkin transformations and the default covariance model with optionally excluded random effects, the variances of the degradation parameters are estimated using mean_degparms, with testing of untransformed log parameters for significant difference from zero. If not using mkin transformations or a custom covariance model, the default initialisation of saemix::saemixModel is used for omega.init.
covariates: A data frame with covariate data for use in 'covariate_models', with dataset names as row names.
covariate_models: A list containing linear model formulas with one explanatory variable, i.e. of the type 'parameter ~ covariate'. Covariates must be available in the 'covariates' data frame.
no_random_effect: Character vector of degradation parameters for which there should be no variability over the groups. Only used if the covariance model is not explicitly specified.
error.init: Will be passed to saemix::saemixModel().
nbiter.saemix: Convenience option to increase the number of iterations
control: Passed to saemix::saemix.
verbose: Should we print information about created objects of type saemix::SaemixModel and saemix::SaemixData?
quiet: Should we suppress the messages saemix prints at the beginning and the end of the optimisation process?
x: An saem.mmkin object to print
digits: Number of digits to use for printing

Details

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 using mkinfit.

Starting values for the fixed effects (population mean parameters, argument psi0 of saemix::saemixModel() are the mean values of the parameters found using mmkin.

Examples

Run this code

if (FALSE) {
ds <- lapply(experimental_data_for_UBA_2019[6:10],
 function(x) subset(x$data[c("name", "time", "value")]))
names(ds) <- paste("Dataset", 6:10)
f_mmkin_parent_p0_fixed <- mmkin("FOMC", ds,
  state.ini = c(parent = 100), fixed_initials = "parent", quiet = TRUE)
f_saem_p0_fixed <- saem(f_mmkin_parent_p0_fixed)

f_mmkin_parent <- mmkin(c("SFO", "FOMC", "DFOP"), ds, quiet = TRUE)
f_saem_sfo <- saem(f_mmkin_parent["SFO", ])
f_saem_fomc <- saem(f_mmkin_parent["FOMC", ])
f_saem_dfop <- saem(f_mmkin_parent["DFOP", ])
anova(f_saem_sfo, f_saem_fomc, f_saem_dfop)
anova(f_saem_sfo, f_saem_dfop, test = TRUE)
illparms(f_saem_dfop)
f_saem_dfop_red <- update(f_saem_dfop, no_random_effect = "g_qlogis")
anova(f_saem_dfop, f_saem_dfop_red, test = TRUE)

anova(f_saem_sfo, f_saem_fomc, f_saem_dfop)
# The returned saem.mmkin object contains an SaemixObject, therefore we can use
# functions from saemix
library(saemix)
compare.saemix(f_saem_sfo$so, f_saem_fomc$so, f_saem_dfop$so)
plot(f_saem_fomc$so, plot.type = "convergence")
plot(f_saem_fomc$so, plot.type = "individual.fit")
plot(f_saem_fomc$so, plot.type = "npde")
plot(f_saem_fomc$so, plot.type = "vpc")

f_mmkin_parent_tc <- update(f_mmkin_parent, error_model = "tc")
f_saem_fomc_tc <- saem(f_mmkin_parent_tc["FOMC", ])
anova(f_saem_fomc, f_saem_fomc_tc, test = TRUE)

sfo_sfo <- mkinmod(parent = mkinsub("SFO", "A1"),
  A1 = mkinsub("SFO"))
fomc_sfo <- mkinmod(parent = mkinsub("FOMC", "A1"),
  A1 = mkinsub("SFO"))
dfop_sfo <- mkinmod(parent = mkinsub("DFOP", "A1"),
  A1 = mkinsub("SFO"))
# The following fit uses analytical solutions for SFO-SFO and DFOP-SFO,
# and compiled ODEs for FOMC that are much slower
f_mmkin <- mmkin(list(
    "SFO-SFO" = sfo_sfo, "FOMC-SFO" = fomc_sfo, "DFOP-SFO" = dfop_sfo),
  ds, quiet = TRUE)
# saem fits of SFO-SFO and DFOP-SFO to these data take about five seconds
# each on this system, as we use analytical solutions written for saemix.
# When using the analytical solutions written for mkin this took around
# four minutes
f_saem_sfo_sfo <- saem(f_mmkin["SFO-SFO", ])
f_saem_dfop_sfo <- saem(f_mmkin["DFOP-SFO", ])
# We can use print, plot and summary methods to check the results
print(f_saem_dfop_sfo)
plot(f_saem_dfop_sfo)
summary(f_saem_dfop_sfo, data = TRUE)

# The following takes about 6 minutes
f_saem_dfop_sfo_deSolve <- saem(f_mmkin["DFOP-SFO", ], solution_type = "deSolve",
  nbiter.saemix = c(200, 80))

#anova(
#  f_saem_dfop_sfo,
#  f_saem_dfop_sfo_deSolve))

# If the model supports it, we can also use eigenvalue based solutions, which
# take a similar amount of time
#f_saem_sfo_sfo_eigen <- saem(f_mmkin["SFO-SFO", ], solution_type = "eigen",
#  control = list(nbiter.saemix = c(200, 80), nbdisplay = 10))
}