Overview (COXREG command)
COXREG applies
Cox proportional hazards regression to analysis of survival times—that
is, the length of time before the occurrence of an event. COXREG supports continuous and categorical
independent variables (covariates), which can be time dependent. Unlike SURVIVAL and KM, which compare only distinct subgroups of cases, COXREG provides an easy way of considering
differences in subgroups as well as analyzing effects of a set of
covariates.
Options
Processing of Independent Variables. You
can specify which of the independent variables are categorical with
the CATEGORICAL subcommand and
control treatment of these variables with the CONTRAST subcommand. You can select one of seven methods
for entering independent variables into the model using the METHOD subcommand. You can also indicate
interaction terms using the keyword BY between variable names on either the VARIABLES subcommand or the METHOD subcommand.
Specifying Termination
and Model-Building Criteria. You can specify the criteria
for termination of iteration and control variable entry and removal
with the CRITERIA subcommand.
Adding New Variables
to Active Dataset. You can use the SAVE subcommand to save the cumulative survival, standard
error, cumulative hazard, log-minus-log-of-survival function, residuals,
XBeta, and, wherever available, partial residuals and DfBeta.
Output. You can print optional output using the PRINT subcommand, suppress or request plots with the PLOT subcommand, and, with the OUTFILE subcommand, write data files containing
coefficients from the final model or a survival table. When only time-constant
covariates are used, you can use the PATTERN subcommand to specify a pattern of covariate values in addition
to the covariate means to use for the plots and the survival table.
Basic Specification
- The minimum
specification on
COXREGis a dependent variable with theSTATUSsubcommand. - To analyze the influence of time-constant
covariates on the survival times, the minimum specification requires
either the
WITHkeyword followed by at least one covariate (independent variable) on theVARIABLESsubcommand or aMETHODsubcommand with at least one independent variable. - To analyze the influence of time-dependent covariates on the survival times, the
TIME PROGRAMcommand and transformation language are required to define the functions for the time-dependent covariates.
Subcommand Order
- The
VARIABLESsubcommand must be specified first; the subcommand keyword is optional. - Remaining subcommands can be named in any order.
Syntax Rules
- Only one dependent
variable can be specified for each
COXREGcommand. - Any number of covariates (independent variables) can be specified. The dependent variable cannot appear on the covariate list.
- The covariate list is
required if any of the
METHODsubcommands are used without a variable list or if theMETHODsubcommand is not used. - Only one status variable can be specified on the
STATUSsubcommand. If multipleSTATUSsubcommands are specified, only the last specification is in effect. - You can use
the
BYkeyword to specify interaction between covariates.
Operations
-
TIME PROGRAMcomputes the values for time-dependent covariates. See the topic TIME PROGRAM for more information. -
COXREGreplaces covariates specified onCATEGORICALwith sets of contrast variables. In stepwise analyses, the set of contrast variables associated with one categorical variable is entered or removed from the model as a block. - Covariates are screened to detect and eliminate redundancies.
-
COXREGdeletes all cases that have negative values for the dependent variable.
Limitations
- Only one dependent variable is allowed.
- Maximum 100 covariates in a single interaction term.
- Maximum 35 levels
for a
BYvariable onPATTERN.