04-24-2013 07:24 PM
I am using PROC SURVEYLOGISTIC to model survey data with only 11 degress of freedom. The procedure applies the Wald F-test. However, the survey methodologists recommend the Satterthwaite adjusted F-test when analyzing this data rather than the adjusted Wald-F and their variations because the latter have been found to be unstable when the degrees of freedom are limited as is the case with this survey data. I don't seem to be able to find an option for specifying the type of F-test to be used for estimating p-values. I know this option is available with SUDAAN software. Does anyone have a recommendation how to produce Satterthwaite adjusted F-test statistics for a logistic model using data with bootstrap replicate weights?