Flexible Bayesian Human Fecundity Models.

Area of Science:

• Reproductive Epidemiology
• Biostatistics
• Human Fecundity Research

Background:

• Human fecundity, crucial for reproductive health, depends on biological capacity and behavioral factors.
• Existing Bayesian hierarchical models use fixed link functions to estimate conception probabilities, potentially causing bias.

Purpose of the Study:

• To propose a generalized class of models for human fecundity by relaxing the fixed link function assumption.
• To improve the accuracy of estimating conception probabilities by allowing flexible link function selection.

Main Methods:

• Developed a general class of models within the generalized nonlinear model framework.
• Utilized the Oxford Conception Study (OCS) dataset for model illustration and fit assessment.
• Employed Markov chain Monte Carlo (MCMC) for Bayesian computations and information criteria (DIC, PSML) for link function selection.

Main Results:

• Demonstrated the utility and improved fit of the proposed flexible models using OCS data.
• Confirmed that the choice of link function significantly impacts the estimation of conception probabilities.
• The proposed models offer a more robust approach to modeling human fecundity.

Conclusions:

• Emphasizes the critical importance of carefully selecting the link function in conception probability models.
• The proposed flexible modeling approach provides more reliable estimates for human fecundity.
• This work contributes to more accurate epidemiological and clinical understanding of human reproduction.