At least for certain data, such as mortgage loan default data, the data is preferably separated into two components, a binary variable indicating whether the event was observed or not, and a time observed variable which is either the time to the event, or if the event has not been observed, the time until the observation was censored.