Methods for assessing difference between groups in change when initial measurements is subject to intra-individual...

Area of Science:

• Biostatistics
• Statistical Modeling
• Clinical Trial Analysis

Background:

• Adjusting for initial values in statistical analysis is crucial for comparing group changes.
• Intra-individual variation complicates accurate estimation of between-group differences.
• Ordinary Least Squares (OLS) can introduce asymptotic bias in such analyses.

Purpose of the Study:

• To estimate the asymptotic bias of OLS when accounting for intra-individual variation.
• To develop corrected estimators and their variances for unbiased group comparisons.
• To explore alternative methods using OLS on transformed data for unbiased estimation.

Main Methods:

• Estimation of asymptotic bias in OLS for adjusted between-group differences.
• Development of explicit formulae using OLS estimates, initial value differences, and intra-individual variation measures.
• Application of OLS on transformed data with conditional Stein estimates of true values.

Main Results:

• Quantification of asymptotic bias introduced by intra-individual variation in OLS analyses.
• Provision of a corrected estimator and its variance for improved accuracy.
• Demonstration of unbiased estimation through OLS on transformed data.

Conclusions:

• Standard OLS methods are biased when intra-individual variation is present and initial values are adjusted for.
• Corrected estimators and OLS on transformed data provide unbiased estimates of between-group differences.
• The findings are applicable to both observational studies and clinical trials.