Online Program

Graphing and Testing Interval Censored Time to Event Data: Practical Aspects of Using the interval R package

Keywords: survival analysis; Kaplan-Meier; R package; logrank tests

This talk is about basic analyses for interval censored responses, which are responses where the event time is not observed exactly but only known to be between two assessment times. For example, in an HIV vaccine trial, time from randomization to the first time that HIV would be detectable in a blood sample is an interval censored response, since the time to the event is bracketed between the last negative and the first positive HIV blood sample (if it exists). The nonparametric maximum likelihood estimator of survival is the generalization of the Kaplan-Meier curve, and allows plotting the survival distribution with minimal assumptions. A generalization of the logrank test may be applied to test for differences in distributions between treatment groups. Both these methods are available in the interval R package. Since the generalized logrank test is not available in the current SAS/STAT, I use this talk to discuss R packages in general, and how one decides whether to trust a particular R package. As the author of the interval package I will offer some inside views on the development of the interval package and the steps that were taken to ensure validity of the package.