High–Dimensional Sparse Matched Case–Control and Case–Crossover Data: A Review of Recent Works, Description of an R Tool and an Illustration of the Use in Epidemiological Studies
Abstract
The conditional logistic regression model is the standard tool for the analysis of epidemiological studies in which one or more cases (the event of interest), are matched with one or more controls (not showing the event). These situations arise, for example, in matched case–control and case–crossover studies. In sparse and high-dimensional settings, penalized methods, such as the Lasso, have emerged as an alternative to conventional estimation and variable selection procedures. We describe the R package clogitLasso, which brings together algorithms to estimate parameters of conditional logistic models using sparsity-inducing penalties. Most individually matched designs are covered, and, beside Lasso, Elastic Net, adaptive Lasso and bootstrapped versions are available. Different criteria for choosing the regularization term are implemented, accounting for the dependency of data. Finally, stability is assessed by resampling methods. We previously review the recent works pertaining to clogitLasso. We also report the use in exploratory analysis of a large pharmacoepidemiological study.