Outlier robust nonlinear mixed model estimation

Area of Science:

• Statistics
• Biostatistics
• Data Analysis

Background:

• Standard nonlinear mixed models are sensitive to outliers, which can significantly distort parameter estimates and standard errors.
• Misleading inferences arise from parameter distortions, impacting the reliability of statistical analyses.
• Aberrant observations, whether individual points or entire clusters, pose a challenge in mixed-model data.

Purpose of the Study:

• To introduce a novel outlier robust method for nonlinear mixed models.
• To address the distortion of parameter estimates and variance components caused by aberrant data.
• To provide a reliable alternative to standard estimation techniques in the presence of outliers.

Main Methods:

• A linearization-based robust method is proposed for estimating fixed effects parameters and variance components.
• The method is designed to mitigate the impact of aberrant observations within clusters or entire clusters.
• Comparative analysis with nonrobust methods using real-world data.

Main Results:

• The proposed robust method yields more accurate parameter estimates compared to standard nonrobust methods.
• Variance component estimation is also improved, leading to more reliable standard errors.
• Demonstrated effectiveness using a four-parameter logistic model with bioassay data.

Conclusions:

• The outlier robust linearization method offers a significant improvement for nonlinear mixed model analyses.
• This approach enhances the accuracy and reliability of statistical inferences in the presence of outliers.
• The method provides a valuable tool for researchers dealing with potentially contaminated data in mixed-effects modeling.