ABSTRACT
Interval-Censored Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 12.7 Statistical Analysis of Other Interval-Censored Data . . . . . . . . . . . . . . . . . . 322 12.8 Software: An Example and Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . 323 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
Interval-censored time-to-event or failure time data occur inmany areas especially in cancer oncology and biopharmaceutics. Their analysis has attracted a great deal of attention over last two decades or so. For example, two books have been published on the topic (Sun, 2006; Chen et al., 2012). In addition, a couple of review papers have been published including Gómez et al. (2009) and Zhang and Sun (2010). In the context of failure time data, interval censoring means that the failure time variable of interest is observed or known only to lie within some intervals or windows instead of being observed exactly (Finkelstein, 1986; Kalbfleisch and Prentice, 2002; Sun, 2006). If the interval includes only or reduces to a single time point, one obtains the exact failure time.
