PATTERN Subcommand (CSCOXREG command)

PATTERN specifies the pattern of predictor values to be used for requested plots on the PLOT subcommand and the exported survival file on the OUTFILE subcommand. PATTERN cannot be used when time-dependent predictors calculated by TIME PROGRAM are included in the model.

  • A value must be specified for each variable specified on PATTERN.
  • Covariates that are included in the model but not named on PATTERN are evaluated at their means.
  • Factors that are included in the model but not named on PATTERN are evaluated at the reference category.
  • You can request separate lines for each category of a factor that is in the model. Specify the name of the categorical variable after the keyword BY. The BY variable must be a categorical variable. You cannot specify a value for the BY variable.
  • Multiple PATTERN subcommands can be specified. CSCOXREG produces a set of requested plots for each specified pattern.
  • Piecewise constant predictor paths are also allowed. The path is specified by endtime(valuelist) varname(valuelist) varname(value)…. If varname(valuelist) is used, the length of valuelist must be the same as that for endtime. The varname(value) means that the value of the variable is constant over time.

Example

CSCOXREG t by A with x1 x2
/VARIABLES STATUS=dead(1)
/PLAN FILE=’c:\survey\myfile.csplan’
/PATTERN x1(0.1) x2(3) A(3)
/PLOT SURVIVAL.
  • Predictor pattern x1 = 0.1, x2 = 3, A = 3 is specified by PATTERN.
  • The survival function is plotted for the specified pattern.

Example: Piecewise constant predictor path

CSCOXREG t1 t2 by A with x1 x2
/VARIABLES STATUS=dead(1)
/PLAN FILE=’c:\survey\myfile.csplan’.
/PATTERN t2(10 20 30 50) x1(1 1.2 1.7 1.9) x2(3) BY A
/OUTFILE SURVIVAL='surv.sav'.
  • Two time variables are specified on the CSCOXREG variable list.
  • PATTERN defines the following predictor paths for x1 and x2.
    Table 1. Predictor paths for x1 and x2
    starttime endtime x1 x2
    0 10 1.0 3
    10 20 1.2 3
    20 30 1.7 3
    30 50 1.9 3
  • PATTERN, through BY A, also specifies that each category of factor A is considered separately. Combining different categories of A with the paths for x1 and x2, the total number of paths considered here actually equals the number of categories of A.
  • The survival table for the specified predictor paths are calculated and written to the file surv.sav.