CALCULATE Subcommand (SURVIVAL command)

CALCULATE controls the comparisons of survival for subgroups specified on the COMPARE subcommand.

  • The minimum specification is the subcommand keyword by itself. EXACT is the default.
  • Only one of the keywords EXACT, APPROXIMATE, and CONDITIONAL can be specified. If more than one keyword is used, only one is in effect. The order of precedence is APPROXIMATE, CONDITIONAL, and EXACT.
  • The keywords PAIRWISE and COMPARE can be used with any of the EXACT, APPROXIMATE, or CONDITIONAL keywords.
  • If CALCULATE is used without the COMPARE subcommand, CALCULATE is ignored. However, if the keyword COMPARE is specified on CALCULATE and the COMPARE subcommand is omitted, an error message is generated.
  • Data can be entered into SURVIVAL for each individual case or aggregated for all cases in an interval. The way in which data are entered determines whether an exact or an approximate comparison is most appropriate. See the topic Using Aggregated Data (SURVIVAL command) for more information. .

EXACT. Calculate exact comparisons. This is the default. You can obtain exact comparisons based on the survival experience of each observation with individual data. While this method is the most accurate, it requires that all of the data be in memory simultaneously. Thus, exact comparisons may be impractical for large samples. It is also inappropriate when individual data are not available and data aggregated by interval must be used.

APPROXIMATE. Calculate approximate comparisons only. Approximate comparisons are appropriate for aggregated data. The approximate-comparison approach assumes that all events occur at the midpoint of the interval. With exact comparisons, some of these midpoint ties can be resolved. However, if interval widths are not too great, the difference between exact and approximate comparisons should be small.

CONDITIONAL. Calculate approximate comparisons if memory is insufficient. Approximate comparisons are produced only if there is insufficient memory available for exact comparisons.

PAIRWISE. Perform pairwise comparisons. Comparisons of all pairs of values of the first-order control variable are produced along with the overall comparison.

COMPARE. Produce comparisons only. Survival tables specified on the TABLE subcommand are not computed, and requests for plots are ignored. This allows all available workspace to be used for comparisons. The WRITE subcommand cannot be used when this specification is in effect.

Example

SURVIVAL TABLE = MOSFREE BY TREATMNT(1,3)
  /STATUS = PRISON (1) FOR MOSFREE
  /INTERVAL = THRU 24 BY 3
  /COMPARE /CALCULATE = PAIRWISE.
  • PAIRWISE on CALCULATE computes the Wilcoxon (Gehan) statistic, degrees of freedom, and observed significance levels for each pair of values of TREATMNT, as well as for an overall comparison of survival across all three TREATMNT subgroups: group 1 with group 2, group 1 with group 3, and group 2 with group 3.
  • All comparisons are exact comparisons.

Example

SURVIVAL TABLE = MOSFREE BY TREATMNT(1,3)
  /STATUS = PRISON (1) FOR MOSFREE
  /INTERVAL = THRU 24 BY 3
  /COMPARE /CALCULATE = APPROXIMATE COMPARE.
  • APPROXIMATE on CALCULATE computes the Wilcoxon (Gehan) statistic, degrees of freedom, and probability for the overall comparison of survival across all three TREATMNT subgroups using the approximate method.
  • Because the keyword COMPARE is specified on CALCULATE, survival tables are not computed.