1. Raue, Bioinformatics 31(21), 2015↩︎

  2. Kaschek, dMod, 2016 (https://cran.r-project.org/package=dMod)↩︎

Create a GPF object

It depends on the arguments how the GPF object will be created:

  • if estimates is missing or set to NULL and if filename is a character string pointing to an existing file, an attempt is made to load the GPF object from the file via load_GPF(filename). In this case, the argument uncertainty_correlation is neglected.

  • if estimates is specified and not NULL, then the GPF object is created from the estimates and uncertainty_correlation arguments, and the argument filename is neglected.

GPF(filename, estimates, uncertainty_correlation = data.frame())



Missing, NULL or a character string denoting a path to a GPF-file. This argument is only considered if estimates is missing or set to NULL.


a data.frame with parameter estimates as specified in General Parameter Format.


a data.frame with parameter uncertainty correlation estimates, as specified in General Parameter Format.


A GPF object or the function stops with an error message, if no such object could be created.


This function does not generate a GPF from a project estimates directory - please, check generate_GPFFromIQRnlmeProject for that purpose.

In the estimates-sheet, the following columns are required:

  • PARAMETER: Names of parameters. c("ka", "beta_ka(WT0)", "corr(ka, CL)")

  • TYPE: Use of paramter. Not used internally but useful for the overview. c("MODEL PARAMETER", "COVARIATE")

  • VALUE: Value of parameter in linear scale

  • VALUE.RSE.PERCENT: sigma/VALUE \* 100

  • IIV: Omega (standard deviation) of random effects in "linear scale". See details for some practical examples.


  • TRANSFORMATION: One of c("N", "L", "G", NA). Can be NA only for corr, beta and error parameters

  • UNIT: Not used for sampling but highly recommended

  • COV.FORMULA: Formula for covariate transformation

  • NAME: Verbose description of the parameter

  • COV.REFERENCE: Reference value of the covariate

  • COMMENT: Ad libitum

Some practical examples you might encounter

  1. You have parameter c(ka = 2) with transformation "L" and the corresponding diagonal entry in your random effects vcov matrix is 0.4.

    • What do you put in IIV? Answer: IIV = sqrt(0.4).

    • How does the final value of an indparam look? Answer: The final indparam = exp(log(2) + rnorm(1,0,sqrt(0.4)))


Daniel Lill (IntiQuan), Venelin Mitov (IntiQuan)


if (FALSE) {
m <- IQRmodel(system.file("examples/GPFExamples/model_1cpt_linear_abs1.txt", package = "IQRtools"))
xls <- generate_GPFfromIQRmodel(m)
est <- xls$estimates
uc <- xls$uncertainty_correlation

# Create a GPF without uncertainty correlation matrix
gpf1 <- GPF(estimates = est)