Related Experiment Videos
Modeling conception as an aggregated Bernoulli outcome with latent variables via the EM algorithm
1Statistics and Biomathematics Branch, National Institute of Environmental Health Sciences, Research Triangle Park, North Carolina 27709, USA.
Biometrics
|September 1, 1996
Summary
This study introduces a novel statistical method for analyzing aggregated Bernoulli outcomes, particularly in reproductive epidemiology. The approach models daily conception probabilities and covariate effects, accounting for multiple cycles per couple.
Area of Science:
- Biostatistics
- Epidemiology
- Reproductive Health
Background:
- Analyzing aggregated Bernoulli outcomes presents challenges, especially in reproductive studies where conception is the primary outcome.
- Existing methods may not adequately address the complexities of daily conception probabilities and the influence of covariates over time.
- The dependency among reproductive cycles within couples requires specialized statistical adjustments.
Purpose of the Study:
- To develop a general statistical method for analyzing aggregated Bernoulli outcomes in epidemiological research.
- To model the daily probability of conception, incorporating latent outcomes and covariate effects.
- To address the issue of correlated outcomes from multiple menstrual cycles within the same couple.
Main Methods:
- An Expectation-Maximization (EM) algorithm was developed to maximize the observed-data pseudo-likelihood.
- The method models unobservable latent outcomes linked to daily intercourse events within a menstrual cycle.
- A generalized estimating equation (GEE) approach was employed to adjust for within-couple dependency across multiple cycles.
Main Results:
- The proposed EM algorithm effectively analyzes aggregated Bernoulli outcomes, providing insights into conception probabilities.
- The method allows for the flexible modeling of covariate effects on both cycle viability and daily conception probabilities.
- GEE adjustments successfully account for the dependency introduced by multiple cycles per couple in prospective studies.
Conclusions:
- The developed statistical framework offers a robust approach for analyzing aggregated Bernoulli outcomes in reproductive epidemiology.
- This method facilitates the investigation of covariates influencing conception probability, distinguishing between long-term and transient effects.
- The approach is broadly applicable to any scenario involving aggregated Bernoulli trials influenced by a susceptibility factor and exhibiting intra-subject dependency.