In sebastianueckert/vaplot: Variability Attribution Plots for Nonlinear Mixed Effect Models

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

The vaplot package creates variability attribution plots for non-linear mixed effect models.

library(vaplot)

lst_file <- system.file("testdata", "derivatives_run4.lst", package = "vaplot")

# read NM results & prepare data structure
input <- prepare_va_nm(lst_file)
omega <- input$omega
omega[2,1] <- omega[2,1] <- 0.08
input2 <- update_omega(input, omega)
# calculate variability attribution
va_res <- compute_va(input, facets = "DVID")
va_res2 <- compute_va(input2, facets = "DVID")
plot_va_compare(va_res, va_res2, smooth = T)
# generate va plot
plot_va(va_res, smooth = T)

FREM Models

frem_path <- system.file("testdata", "frem_pheno", package = "vaplot")

# read FREM results & prepare data structure
input <- prepare_va_frem(frem_path)

# calculate variability attribution

va_res <- compute_va(input)
# generate va plot
plot_va(va_res, coloring = coloring_highlight_covs)

Models in R

In addition to NONMEM, the package also supports NLME models written as a simple R function.

my_model <- function(time, theta, eta, eps, z){
  v <- theta[1]*exp(eta[1])
  cl <- theta[2]*exp(eta[2])
  ka <- theta[3]*exp(eta[3])
  k <- cl/v
  conc <- z["dose"]/v*ka/(ka-k)*(exp(-k*time)-exp(-ka*time))
  y <- conc * (1  + eps[1])
  return(y)
}

theta <- c(3, 5, 0.1)
eta <- c(0,0,0)
z <- c(dose = 10)
time <- seq(0, 20, length.out = 40)
plot(time, my_model(time, theta, eta, 0, z), type = "l")

omega <- diag(c(0.2,0.1,0.5), 3, 3)
sigma <- matrix(.05, 1, 1)

input <- prepare_va_fun(my_model, theta, omega, sigma, time, z)
va_res <- compute_va(input)
plot_va(va_res)

sebastianueckert/vaplot documentation built on Sept. 1, 2021, 11:24 a.m.