New tests for trend in time censored recurrent event data

Abstract

We consider testing for trend in recurrent event data. More precisely, for such data we consider testing of the null hypothesis of data coming from a renewal process. The new tests are essentially obtained by considering appropriate integrated versions of classical trend tests. Moreover, adaptive versions of earlier considered tests versus non-monotonic alternatives, like bathtub trend, are suggested. A simulation study shows that the new tests have favorable properties and sometimes outperform classical tests. Examples with real data are also considered.