The LIFETEST Procedure

TIME Statement

  • TIME variable <*censor(list)> </ option>;

The TIME statement is required. It is used to indicate the failure time variable, where variable is the name of the failure time variable that can be optionally followed by an asterisk, the name of the censoring variable, and a parenthetical list of values that correspond to right-censoring. The censoring values should be numeric, nonmissing values. For example, the following statement identifies the variable T as containing the observed failure times (event or censored): 

time T*Status(0,2);

If the variable Status has the value 0 or 2, the corresponding value of T is a right-censored value.

You can specify the following option after a slash (/):

FAILCODE<=number-list>
EVENTCODE<=number-list>

stipulates a competing-risks analysis, which consists of estimating cumulative incidence functions and computing Gray’s test (Gray 1988) for testing the homogeneity of two or more cumulative incidence functions. You specify a number that represents the event of interest after the equal sign. For example: 


proc lifetest;
   time T*Status(0) / failcode=1;
run;

For this specification, PROC LIFETEST regards a Status value of 1 as the event of interest, a value of 0 as a censored observation indicator, and all other values as competing events.

You can specify a list of values after the equal sign. PROC LIFETEST performs a separate competing-risks analysis for each value, regarding it as representing the event of interest. For example: 


proc lifetest;
   time T*Status(0) / failcode=1 2;
run;

This specification produces two analyses, one for FAILCODE=1 and the other for FAILCODE=2.

If you specify the FAILCODE option without the equal sign, PROC LIFETEST produces a separate analysis for each distinct event value. Consider a data set with an event indicator variable Status that assumes four distinct values, 0, 1, 2, and 3, where Status=0 represents observations that are censored and Status=1, Status=2, and Status=3 represent three different causes of failure. Consider the following statements: 


proc lifetest;
   time T*Status(0) / failcode;
run;

PROC LIFETEST produces three separate competing-risks analyses: one uses Status=1 as the failure cause of interest, one uses Status=2 as the failure cause of interest, and one uses Status=3 as the failure cause of interest. This specification is convenient for an exploratory analysis when there is no predetermined failure cause of interest.