Event Times

A crucial concept in survival analysis is the time elapsed between some landmark and a particular event of interest. As an example, say you're running a clinical trial to investigate the efficacy of a new anticonvulsant. You may be interested in the time from the start of therapy to the first epileptic seizure for each patient. But if a patient dies or otherwise goes off study before they have a seizure, you'll assume that a seizure would have occurred eventually, but you don't know when exactly. In this case the event time is right censored; the true event time is unknown, all you know is that it exceeds the observed time.

A dedicated type is provided to conveniently store right censored data.

Survival.EventTimeType
EventTime{T}

Immutable object containing the real-valued time to an event as well as an indicator of whether the time corresponds to an observed event (true) or right censoring (false).

source

Summarizing Event Times

Given times to an event of interest and indications of whether the observations are right censored, we can construct a table of the unique times along with the number of events of interest, the number of censored observations, and the size of the risk set at each time. This information is used for computing other estimates, e.g. of the survivor and cumulative hazard functions.

Survival.EventTableType
EventTable{T}

Immutable object summarizing the unique observed event times, including the number of events, the number of censored observations, and the number remaining at risk for each unique time.

This type implements the Tables.jl interface for tables, which means that EventTable objects can be seamlessly converted to other tabular types such as DataFrames.

EventTable(eventtimes)

Construct an EventTable from an array of EventTime values.

EventTable(time, status)

Construct an EventTable from an array of time values and an array of event status indicators.

source