Bivariate modelling of clustered continuous and ordered categorical outcomes

Area of Science:

• Biostatistics
• Developmental Toxicology
• Statistical Modeling

Background:

• Biomedical research frequently involves simultaneous observation of continuous and ordered categorical outcomes.
• Multivariate analysis of such mixed-type data presents significant challenges due to differing data structures.

Purpose of the Study:

• To develop a statistical model for the joint distribution of bivariate continuous and ordinal outcomes.
• To extend the model to accommodate clustering within bivariate outcomes.
• To facilitate parameter estimation using a convenient factorization and estimating equations.

Main Methods:

• Utilized latent variable concepts applied to a multivariate normal distribution to model joint distributions.
• Extended the model to incorporate clustering of bivariate outcomes.
• Parameterized the joint distribution as a product of a random effects model (continuous) and a correlated cumulative probit model (ordinal).

Main Results:

• Successfully constructed a flexible model for analyzing mixed-type bivariate outcomes.
• Demonstrated a practical factorization for parameter estimation via estimating equations.
• Illustrated the model's application using foetal weight and malformation data from a developmental toxicity study.

Conclusions:

• The proposed model effectively addresses the complexities of analyzing joint continuous and ordinal data in biomedical research.
• The factorization simplifies parameter estimation, making the approach computationally feasible.
• This methodology offers a valuable tool for understanding outcomes in developmental toxicity and similar research areas.