Title: Variable selection for the estimation of marginal hazard ratios in high-dimensional data
Abstract: Hazard ratios are important measures frequently reported in time-to-event and epidemiological studies in the investigation of treatment effects. In observational studies, through the use of propensity score weights in combination with the Cox proportional hazards model, a possible measure to be estimated is the marginal hazard ratio (MHR). The method for for the estimation of the MHR is analogous to those used to estimated causal parameters, like the average treatment effect (ATE). However, unlike previous research in causal inference, no studies have been conducted in relation to the MHR estimation considering high dimensional data. This paper aims to cover this gap through a simulation study considering variable selection methods from causal inference in combination with recent multi-robust approaches developed for MHR estimation. Results from the simulation study indicate that double selection (DS) is an adequate method when estimating the MHR, but it can be further improved with the use of a multi-robust approach on the set of propensity score models obtained during the double selection process.