On the heterogeneity of fecundability

Area of Science:

• Biostatistics
• Reproductive Epidemiology
• Statistical Modeling

Background:

• Couple fecundability exhibits significant heterogeneity, complicating accurate modeling.
• Existing statistical methods may not fully capture the complexities of fertility data, especially with censored observations.

Purpose of the Study:

• To develop a flexible statistical framework for modeling heterogeneity in couple fecundability.
• To introduce canonical moments for characterizing mixing distributions and improve estimation accuracy.
• To provide a method applicable to censored data and the identification of sterile subpopulations.

Main Methods:

• A general mixing distribution is proposed to model fecundability variation.
• Canonical moments are introduced as parameters to define the mixing distribution.
• Bootstrap methods are employed for robust standard error estimation.
• The approach modifies moment estimates to ensure support on [0, 1] and reduce bias.

Main Results:

• The proposed method yields a mixing distribution supported on [0, 1], ensuring valid probability estimates.
• A reduction in the downward bias of moment estimates for the number of support points is achieved.
• The technique demonstrates applicability to censored data and the identification of sterile subpopulations.

Conclusions:

• The novel mixing distribution and canonical moments provide a statistically sound and flexible approach to modeling couple fecundability.
• This method enhances the accuracy of fertility parameter estimation and offers practical applications in reproductive health research.