getsigmas: Estimate individual-level residual error from the elimination...

In nlmixr2autoinit: Automatic Generation of Initial Estimates for Population Pharmacokinetic Modeling

Estimate individual-level residual error from the elimination phase

Description

Performs log-linear regression on the elimination phase of a single individual's or one group's pharmacokinetic concentration–time data to estimate additive and proportional residual standard deviations.

Usage

getsigmas(group_df, nlastpoints = 3)

Arguments

group_df	A data frame for a single group (e.g., one subject or dose), containing columns: EVID (event ID), DV (observed concentration), TIME (time after dose), and routeobs (administration route).
nlastpoints	Integer specifying the number of terminal data points used for regression.

Details

Residuals are computed from individual-predicted concentrations (IPRED) and observed concentrations (DV) using the following definitions:

\sigma_{add} = \sqrt{Var(C_{obs} - C_{pred})}

\sigma_{prop} = \sqrt{Var\left(\frac{C_{obs}}{C_{pred}} - 1\right)}

where C_{obs} is the observed concentration and C_{pred} is the model-predicted concentration obtained by back-transformation of the log-linear regression. The additive residual standard deviation (\sigma_{add}) and proportional residual standard deviation (\sigma_{prop}) are calculated per individual.

Value

A tibble with the following columns:

• intercept: Intercept of the log-linear regression line

• slope: Estimate of the terminal elimination rate constant

• residual_sd_additive: Standard deviation of additive residuals

• residual_sd_proportional: Standard deviation of proportional residuals

Author(s)

Zhonghui Huang

Examples

dat <- Bolus_1CPT
dat <- processData(dat)$dat
getsigmas(dat[dat$ID == 1 & dat$dose_number == 1 & dat$resetflag == 1 &
              dat$EVID == 0, ])

nlmixr2autoinit documentation built on Nov. 14, 2025, 1:07 a.m.