18
2.3 Introduction to survival analysis
According to the classical definition (Miller, 1981), survival analysis focuses on the
time until the event of interest occurs. The event of interest should be defined to
serve specific research topics. For example, when studying clinical trials, the event of
interest can be defined as the death of a patient if the purpose is to analyze patients’
survival; or it can be defined as a disease incidence if the purpose is to look into
disease occurrences.
Depending on the research objective, the event of interest could be defined broadly.
It could be the time to learn a skill, or it may not even be a time at all. For instance, it
could be the number of dollars that an insurance company pays to a particular claim.
This characteristic can be particularly useful. It allows us to apply the survival
analysis techniques to other fields, with much broader applications.
2.3.1 Censoring
Censoring is a special feature under the survival analysis setting. When there is an
incomplete observation of the failure time, we call it a censoring. For example, a pa-
tient lost to follow up or withdraw from a clinical trial. Censoring can be categorized
in different ways.
Type I censoring occurs when a study ends when a certain time point is reached.
Type II censoring occurs when a study ends when a certain number of failures
occurs. In both cases, the event of interest is not observed.