Related Experiment Videos
The effect of collinearity on parameter estimates in nonlinear mixed effect models.
1Quintiles Inc., Kansas City, Missouri 64134-0708, USA. pbonate@qkan.quintiles.com
Pharmaceutical Research
|June 1, 1999
Summary
High correlation between predictor variables in population pharmacokinetic models can bias parameter estimates and inflate precision. Researchers should be cautious with correlations above 0.5 to ensure model accuracy.
Area of Science:
- Pharmacokinetics
- Population modeling
- Statistical analysis
Background:
- Population pharmacokinetic (PopPK) models are crucial for understanding drug behavior.
- Identifying significant covariates is essential for accurate PopPK model development.
- Covariate correlations can complicate the selection of relevant predictors.
Purpose of the Study:
- To investigate the impact of predictor variable correlations on covariate selection in PopPK models.
- To assess how increasing covariate correlation affects the accuracy and precision of parameter estimates.
Main Methods:
- Utilized Monte Carlo simulations to generate concentration-time data.
- Simulated data with varying degrees of correlation between two covariates.
- Analyzed data using NONMEM, assessing model performance with and without covariate inclusion.
Main Results:
- Increased covariate correlation led to larger standard errors and biased parameter estimates.
- A covariate correlation of 0.75 resulted in parameter estimates too imprecise for statistical significance.
Conclusions:
- Correlations exceeding 0.5 between simultaneously modeled covariates serve as a warning.
- High correlations can lead to biased parameter accuracy and inflated precision due to ill-conditioning.