A semiparametric empirical likelihood method for data from an outcome-dependent sampling scheme with a continuous...

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Case-control studies are common but lose efficiency with continuous outcomes.
• Outcome-dependent sampling (ODS) offers enhanced efficiency by collecting supplemental samples based on outcome values.
• Dichotomizing continuous outcomes in studies can lead to significant information loss.

Purpose of the Study:

• To introduce a semiparametric empirical likelihood inference procedure for ODS designs with continuous outcomes.
• To evaluate the efficiency and properties of the proposed estimator compared to existing methods.
• To apply the novel method to a real-world environmental epidemiology dataset.

Main Methods:

• Developed a semiparametric empirical likelihood inference procedure treating covariate distribution as a nuisance parameter.
• Utilized ODS by incorporating an overall random sample and supplemental samples based on a continuous outcome.
• Assessed asymptotic normality and Wilks-type properties of the likelihood ratio statistic.

Main Results:

• The proposed semiparametric empirical likelihood estimator demonstrated superior efficiency over conditional likelihood and probability weighted pseudolikelihood estimators.
• ODS designs combined with the new estimator yielded more efficient results than simple random sampling designs of equivalent size.
• Simulations confirmed the asymptotic normality and chi-square distribution of the likelihood ratio statistic under the null hypothesis.

Conclusions:

• The semiparametric empirical likelihood method is an efficient tool for analyzing data from ODS designs with continuous outcomes.
• ODS designs, when coupled with appropriate statistical methods, can significantly improve the efficiency of epidemiologic studies.
• The method was successfully applied to investigate the association between maternal PCB levels and children's IQ in the Collaborative Perinatal Project study.