A semiparametric empirical likelihood method for biased sampling schemes with auxiliary covariates

Area of Science:

• Biostatistics
• Epidemiology
• Genomics

Background:

• Epidemiologic studies often use complex sampling schemes to oversample specific subpopulations.
• Accurate statistical inference is challenging when sampling probabilities are unknown.
• Understanding associations between biomarkers and treatment response is crucial in precision medicine.

Purpose of the Study:

• To develop a semiparametric inference procedure for two-component sampling schemes in epidemiologic studies.
• To address challenges of nonidentifiable sampling probabilities in regression analysis.
• To assess the association between epidermal growth factor receptor (EGFR) mutation levels and antitumor response in non-small cell lung cancer (NSCLC) patients.

Main Methods:

• Developed a semiparametric inference procedure for data with simple random samples and dependent samples.
• Applied the method within the generalized linear models framework with arbitrary link functions.
• Utilized simulation studies to evaluate the estimator's performance.

Main Results:

• The proposed semiparametric estimator demonstrates favorable small sample properties.
• The method is applicable to both binary and multicategorical outcome data.
• The procedure effectively handles complex sampling designs with nonidentifiable sampling probabilities.

Conclusions:

• The novel statistical method provides a robust approach for analyzing complex epidemiologic data.
• This technique enhances the ability to study biomarker-treatment associations in diseases like NSCLC.
• The findings support improved statistical inference in precision oncology research.