Empirical Likelihood-Based Estimation of the Treatment Effect in a Pretest-Posttest Study

Area of Science:

• Biostatistics
• Medical Research Methodology
• Statistical Inference

Background:

• Pretest-posttest designs are standard for evaluating interventions.
• Existing methods face challenges with missing posttest data and model misspecification.
• There is a need for more efficient and robust statistical inference procedures.

Purpose of the Study:

• To propose a semiparametric empirical likelihood (EL) estimation procedure for pretest-posttest data.
• To incorporate baseline covariate information for enhanced efficiency.
• To provide an asymptotically unbiased estimate of the response distribution and improve treatment effect estimation, particularly for skewed data.

Main Methods:

• Developed a semiparametric estimation procedure using empirical likelihood (EL).
• Incorporated baseline covariate information into the EL framework.
• Evaluated the method's performance through simulation studies and analysis of clinical trial data.

Main Results:

• The proposed EL estimator demonstrates improved efficiency, especially when the working model is misspecified.
• The method yields asymptotically unbiased estimates of the response distribution.
• Simulation studies confirmed the EL-based estimator's superiority over competitors with missing at random data and model misspecification.

Conclusions:

• The empirical likelihood method offers a robust and efficient approach for pretest-posttest data analysis, particularly with missing outcomes.
• This method provides a more appealing estimate of treatment effects for skewed data.
• The findings have implications for improving statistical inference in medical and social science research.