Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
6868685 | Computational Statistics & Data Analysis | 2018 | 15 Pages |
Abstract
Joint analysis of multivariate outcomes composed of mixed data types (continuous, count, binary, survival, etc.) induces special complexity in model specification and analysis. When the scientific question of interest involves a joint effect of covariate(s) of interest on the set of outcome variables, specifying a full probability model may be infeasible, undesirably complex, or computationally intractable. A flexible method to estimate and conduct inference on such joint effects is presented which accounts for correlation among the outcomes without needing to explicitly specify their joint distribution. Simulation studies and an analysis of health care data illustrate the approach and its operating characteristics vis-Ã -vis other methods.
Related Topics
Physical Sciences and Engineering
Computer Science
Computational Theory and Mathematics
Authors
Scott Marchese, Guoqing Diao,