Handling missing birthdates in marginal regression analysis with recurrent events

In attempt to provide a practical guide to handling missing birthdate information, this paper examines the strategy proposed by Hu and Rosychuk (2016) for estimating age-varying effects in a marginal regression analysis of recurrent event times. We conduct empirical studies based on the same dataset that motivated Hu and Rosychuk's research and explore how analysis outcomes differ when using different distributions for missing birthdates in situations with different sample sizes. Our studies show that Hu and Rosychuk's assumption of uniformly-distributed birthdates is an appropriate and computationally efficient solution to restricted birthdate information with a reasonably large sample.