Estimation of group means when adjusting for covariates in generalized linear models

Area of Science:

• Biostatistics
• Statistical Modeling
• Data Analysis

Background:

• Generalized linear models (GLMs) are standard for analyzing categorical data (binary, count, ordinal).
• Adjusting for covariates in GLMs can enhance estimation efficiency.
• Current GLM software may produce biased group mean estimates by evaluating at the mean covariate, not the population mean response.

Purpose of the Study:

• To develop a novel method for consistently estimating group means in GLMs.
• To provide accurate variance estimation alongside the group mean estimates.
• To address the bias issue in model-based group mean estimation for GLMs.

Main Methods:

• Proposed a new statistical method for group mean estimation in GLMs.
• Incorporated consistent variance estimation with the proposed method.
• Utilized simulation studies to evaluate the performance of the new method.

Main Results:

• The proposed method yields an unbiased estimator for group means in GLMs.
• Simulation results demonstrated correct coverage probabilities with the new method.
• The method was successfully applied to real-world hypoglycemia data from diabetes clinical trials.

Conclusions:

• The developed method offers a statistically sound approach to estimating group means in generalized linear models.
• This overcomes limitations of existing software, providing more accurate estimates for categorical outcome data.
• The findings have practical implications for analyzing clinical trial data, such as in diabetes research.