What is the simplest approach to analysis of recurrence data?

• A. Generalized proportional hazard model.
• B. Standard Cox model.
• C. Parametric exponential model.
• D. Poisson regression.

Answer: (A)

Recurrence data involve multiple events within the same subject, so you analyze them with a framework that extends survival analysis to a counting process. The generalized proportional hazards model keeps the familiar hazard structure but lets each subject have several event times, updating the risk set after every event. In practice you model the hazard as h(t|history) = h0(t) exp(Xβ), where the baseline hazard is left unspecified and time-varying information or event history can be included. This approach is the simplest extension because it uses the Cox-type form you already know, allows multiple failures per subject, and yields robust standard errors to account for the correlation of events within the same person, without forcing strong parametric assumptions about the inter-event times. By contrast, the standard Cox model is primarily for the time to the first event and would require ad hoc adjustments to handle recurrences; a parametric exponential model imposes a strict inter-event time distribution; Poisson regression treats counts over fixed intervals and may ignore the exact timing of events, losing information.