The process of censoring yields data w/ only partial information. The most common example of censoring is *right censoring* in survival analysis, where the time until the event occurred is only known to be longer than some duration because the event had not occurred when the study ended.
The process of censoring yields data with only partial information. Censoring is especially common, and an important topic, in survival analysis. There are three types of censoring:
- Right censoring (the most common) is when a value is only known to be greater than some amount. A typical example is where the time until an event occurred is only known to be longer than some duration because the event had not occurred when the study ended. Right censoring can also occur as a result of dropout or loss to follow up.
- Interval censoring is when a value is only known to lie within some interval. The most common example would be when data have been rounded. Thus, for example, a value of 5 implies the original value was in the interval $[4.5,~5.5)$. In a survival analysis context, it may only be known that a patient's symptoms reemerged between their 2nd and 3rd checkups.
- Left censoring is when a value is only known to be less than some amount. An example might be when the reemergence of symptoms is discovered in the 1st checkup. Note that since durations cannot be $<0$, left censoring can often be considered a special case of interval censoring.
In the planning of survival analysis investigations, studies are sometimes designed to last for a fixed duration and are sometimes designed to last until a fixed number of events have occurred. In the former case, all right censored observations when the study finishes are due to type I censoring, whereas in the latter case, right censored observations are due to type II censoring.