Rsmlx (version 2023.1.5)

testmlx: Statistical tests for model assessment

Description

Perform several statistical tests using the results of a Monolix run to assess the statistical components of the model in use.

Usage

testmlx(
  project = NULL,
  tests = c("covariate", "randomEffect", "correlation", "residual"),
  plot = FALSE,
  adjust = "edf",
  n.sample = NULL
)

Value

a list of data frames and ggplot objects if plot=TRUE

Arguments

project

a Monolix project

tests

a vector of strings: the list of tests to perform among c("covariate","randomEffect","correlation","residual")

plot

FALSE/TRUE display some diagnostic plots associated to the tests (default=FALSE)

adjust

method to take into account the dependency of MCMC sample c("edf","BH") (default="edf")

n.sample

number of samples from the conditional distribution to be used (default = number of available samples in the project)

Details

The tests used are: 1) F-tests (or, equivalently, correlation tests) to evaluate the effect of each covariate on each parameter ("covariate"), 2) correlation tests to assess the correlation structure of the random effects ("correlation"), 3) Shapiro-Wilk and Miao-Gel-Gastwirth tests to assess, respectively the normality and the symmetry of the distribution of the random effects (""randomEffect"), 4) Shapiro-Wilk and Miao-Gel-Gastwirth tests to assess, respectively the normality and the symmetry of the distribution of residual errors ("residual").

By default, the four tests are performed.

When several samples of the conditional distributions are used, two methods are proposed in order to take into the dependance of the samples for the Shapiro-Wilk and Miao-Gel-Gastwirth tests: "edf" computes an effective degrees of freedom, "BH" performs one test per replicates and adjust the smallest p-value using the Benjamini-Hochberg correction.

Examples

# RsmlxDemo2.mlxtran is a Monolix project for modelling the PK of warfarin using a PK model 
# with parameters ka, V, Cl.

#testmlx will perform statistical tests for the different component of the statistical model:
r1 <- testmlx(project="RsmlxDemo2.mlxtran")

#testmlx will perform statistical tests for the covariate model and the correlation model only.
r2 <- testmlx(project="RsmlxDemo2.mlxtran", tests=c("covariate","correlation"))

# See http://monolix.lixoft.com/rsmlx/testmlx/ for detailed examples of use of testmlx
# Download the demo examples here: http://monolix.lixoft.com/rsmlx/installation