SDA 3.5 Documentation for SE COMPARISONS


SE Comparisons - Comparison of Standard Errors with Other Packages


Most of the SDA analysis programs compute standard errors for complex samples. The programs that can handle complex samples are TABLES, MEANS, REGRESS, and LOGIT. Details of the computation of standard errors and related statistics are given in the online help file.

Comparison runs have been done to check the SDA results, and the results from SDA are generally identical or very similar to results from SAS, Stata, SUDAAN, and SPSS. Some of those comparison runs are assembled here, so that SDA users can feel more confident of the results obtained from SDA.

SDA deliberately does a few things differently from one or another of those other programs, and they also differ among themselves on certain points. We will try to clarify those differences and explain why SDA calculates results as it does.

(This document is under construction. We will gradually assemble various comparison runs and make them available here.)

CSM, UC Berkeley
April 12, 2011