Recurrent event data are frequently encountered in public health and medical applications. Existing methods for analysis typically focus on the analysis of event intensities, rates, or counts, with the fundamental time scale either being "calendar time" or, often more naturally, "gap time" (i.e., inter-event time). Using a suitable adaptation of generalized estimating equations (GEE) that accounts for the longitudinal but informatively censored nature of the observed gap times, we propose a methodology for estimating the parameters that index the conditional means and variances of the process gap times. The proposed methodology permits the use of both time-fixed and time-varying covariates, as well as transformations of the gap times, creating a flexible and useful class of methods for analyzing recurrent event data under comparatively weak assumptions. Data from a randomized trial of asthma prevention in young children serves as both motivation and example. This is joint work with David Clement.
Group for Research in Decision Analysis