The SURVEYPHREG Procedure

Degrees of Freedom

PROC SURVEYPHREG uses the degrees of freedom of the variance estimator to obtain t confidence limits and Wald type F tests. PROC SURVEYPHREG computes the degrees of freedom based on the variance estimation method and the sample design. Alternatively, you can specify the degrees of freedom in the DF= option in the MODEL statement.

For Taylor series variance estimation, PROC SURVEYPHREG calculates the degrees of freedom (df) as the number of clusters minus the number of strata. If the CLUSTER statement is not specified, then the procedure treats each observation as a cluster. If the STRATA statement is not specified, then the procedure assumes that all observations are in the same stratum. These numbers are based on the observations included in the analysis. These numbers do not count observations that are excluded from the analysis due to missing values. See the section Missing Values for details. If you specify the MISSING option in the CLASS statement, missing values are treated as valid nonmissing levels and are included in computing degrees of freedom. If you specify the NOMCAR option for Taylor series variance estimation, observations with missing values of the analysis variables are included in computing the degrees of freedom.

If you provide replicate weights with a REPWEIGHTS statement, the degrees of freedom is equal the number of replicates used, which is the number of REPWEIGHTS variables that provide replicate estimates. Alternatively, you can specify the degrees of freedom in the DF= option in the REPWEIGHTS or MODEL statement.

For BRR variance estimation (when you do not use a REPWEIGHTS statement), PROC SURVEYPHREG calculates the degrees of freedom as the number of strata. PROC SURVEYPHREG bases the number of strata on all valid observations in the data set. If some replicate samples are not usable in the sense that these replicate samples cannot be used for parameter estimation (say, for nonconvergence or inestimability), then df equals the minimum of the number of strata and the number of replicates used. Alternatively, you can use the DF= ALLREPS option in the MODEL statement to specify that df equals the number of strata.

For jackknife variance estimation (when you do not use a REPWEIGHTS statement), PROC SURVEYPHREG calculates the degrees of freedom as the number of clusters minus the number of strata. If the CLUSTER statement is not specified, then the procedure treats each observation as a cluster. If the STRATA statement is not specified, then the procedure assumes that all observations are in the same stratum. For jackknife variance estimation, PROC SURVEYPHREG bases the number of strata and clusters on all valid observations in the data set. If some replicate samples are not usable in the sense that these replicate samples cannot be used for parameter estimation (say, for nonconvergence or inestimability), then df equals the number of clusters (or observations if no CLUSTER statement is specified) minus the number of strata (or one in if no STRATA statement is specified) minus the number of replicate samples that are not used. Alternatively, you can use the DF= ALLREPS option in the MODEL statement to specify that df equals the number of clusters minus the number of strata.