Models of varying parametric form in case-referent studies

Area of Science:

• Epidemiology
• Biostatistics
• Statistical Modeling

Background:

• Case-referent studies are crucial for investigating disease causes.
• Analyzing exposure data in these studies requires robust statistical frameworks.
• Existing methods may not fully capture the complexities of disease incidence rates.

Purpose of the Study:

• To develop a statistical approach for analyzing case-referent data.
• To model disease incidence rates using binomial processes.
• To provide a method for comparing different epidemiological models.

Main Methods:

• Treating case and referent counts as binomial processes.
• Modeling binomial parameters to form a joint likelihood function.
• Maximizing the likelihood function to estimate model parameters.
• Utilizing likelihood ratios for model comparison.

Main Results:

• The proposed method allows for the estimation of parameters in additive and multiplicative models of disease incidence.
• Likelihood ratios effectively compare models with similar parameter counts.
• The approach provides a measure of relative hypothesis corroboration.

Conclusions:

• The binomial process framework offers a flexible approach to case-referent data analysis.
• This method enhances the statistical rigor of epidemiological studies.
• It aids in understanding the relative support for different etiological hypotheses.