The ICPHREG procedure compares most closely to the PHREG procedure. Both procedures can fit proportional hazards models. They differ in the types of censored data that they are designed to handle and the forms of the baseline function. Table 63.1 matches the procedures with the types of censored data they can analyze and the form of the baseline function.
Table 63.1: Proportional Hazards Modeling
Data Type |
Baseline |
PROC ICPHREG |
PROC PHREG |
---|---|---|---|
Interval-censored |
Piecewise constant |
Y |
|
Cubic splines |
Y |
||
Unspecified |
|||
Right-censored |
Piecewise constant |
Y |
Y |
Cubic splines |
Y |
||
Unspecified |
Y |
The PHREG procedure deals exclusively with right-censored data, and it mainly adopts a semiparametric approach by leaving the baseline hazard function unspecified. The ICPHREG procedure is specifically designed to handle interval-censored data and offers different options to parameterize the baseline hazard function. You can use the ICPHREG procedure to analyze data that are left-censored, interval-censored, or right-censored. However, if the data to be analyzed contain only exact or right-censored observations, it is recommended that you use the PHREG procedure because it provides specialized methods for dealing with right-censored data. For more information about PROC PHREG, see Chapter 85: The PHREG Procedure.
Both the LIFEREG procedure and the ICPHREG procedure can handle interval-censored data. The LIFEREG procedure focuses on parametric analysis that uses accelerated failure time models, and it can fit only a proportional hazards model that assumes a Weibull baseline hazard function.