The purpose of this method is to check if a certain algorithm for fitting nonlinear hierarchical models (also known as nonlinear mixed-effects models) will reliably yield results that are sufficiently similar to each other, if started with a certain range of reasonable starting parameters. It is inspired by the article on practical identifiabiliy in the frame of nonlinear mixed-effects models by Duchesne et al (2021).
multistart(
object,
n = 50,
cores = if (Sys.info()["sysname"] == "Windows") 1 else parallel::detectCores(),
cluster = NULL,
...
)
# S3 method for saem.mmkin
multistart(object, n = 50, cores = 1, cluster = NULL, ...)
# S3 method for multistart
print(x, ...)The fit object to work with
How many different combinations of starting parameters should be used?
How many fits should be run in parallel (only on posix platforms)?
A cluster as returned by parallel::makeCluster to be used for parallel execution.
Passed to the update function.
The multistart object to print
A list of saem.mmkin objects, with class attributes 'multistart.saem.mmkin' and 'multistart'.
In case the online version of this help page contains error messages in the example code and no plots, this is due to the multistart method not working when called by pkgdown. Please refer to the online vignette in this case.
Duchesne R, Guillemin A, Gandrillon O, Crauste F. Practical identifiability in the frame of nonlinear mixed effects models: the example of the in vitro erythropoiesis. BMC Bioinformatics. 2021 Oct 4;22(1):478. doi: 10.1186/s12859-021-04373-4.
# \dontrun{
library(mkin)
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"]] <- dmta_ds[["Elliot 2"]] <- NULL
f_mmkin <- mmkin("DFOP", dmta_ds, error_model = "tc", cores = 7, quiet = TRUE)
f_saem_full <- saem(f_mmkin)
f_saem_full_multi <- multistart(f_saem_full, n = 16, cores = 16)
#> Warning: all scheduled cores encountered errors in user code
parhist(f_saem_full_multi, lpos = "bottomright")
#> Error in UseMethod("parms", object): no applicable method for 'parms' applied to an object of class "try-error"
f_saem_reduced <- update(f_saem_full, covariance.model = diag(c(1, 1, 0, 1)))
# On Windows, we need to create a cluster first. When working with
# such a cluster, we need to export the mmkin object to the cluster
# nodes, as it is referred to when updating the saem object on the nodes.
library(parallel)
cl <- makePSOCKcluster(12)
clusterExport(cl, "f_mmkin")
#> Error in get(name, envir = envir): object 'f_mmkin' not found
f_saem_reduced_multi <- multistart(f_saem_reduced, n = 16, cluster = cl)
#> Error in checkForRemoteErrors(val): 12 nodes produced errors; first error: object 'f_mmkin' not found
parhist(f_saem_reduced_multi, lpos = "bottomright")
#> Error in parhist(f_saem_reduced_multi, lpos = "bottomright"): object 'f_saem_reduced_multi' not found
#> Warning: calling par(new=TRUE) with no plot
# }