Confidence Interval for Difference Between Means, Separate Variances When sample sizes are unequal, one should always use a separate variances standard error and adjusted degrees of freedom (Some properties of preliminary tests of equality of variances in the two-sample location problem, Journal of General Psychology, 1996, 123, 217-231). This is very easily done with SPSS: T-TEST GROUPS = iup(0 1) /MISSING = ANALYSIS /VARIABLES = Pups /CRITERIA = CI(.95) . Independent Samples Test t-test for Equality of Means t Pups Equal variances as sumed Equal variances not ass umed df Sig. (2-tailed) Mean Difference Std. Error Difference 95% Confidence Interval of the Difference Lower Upper -2.253 19 .036 -7.1923 3.1921 -13.8734 -.5112 -2.784 15.096 .014 -7.1923 2.5831 -12.6951 -1.6896 With SAS (release 9.1 or lower) it is a bit more difficult to get the separate variances confidence interval. Proc TTest provides pooled and separate variances t, but only the pooled confidence interval. If you have SAS release 9.2 or higher, you get both confidence intervals. The output here is from release 9.1. proc ttest; class iup; var pups; run; -----------------------------------------------------------------------------------------------The TTEST Procedure Statistics Variable iup pups N Lower CL Mean Mean Upper CL Mean Lower CL Std Dev Std Dev Upper CL Std Dev Std Err 8 38.357 40.5 42.643 1.6949 2.5635 5.2174 0.9063 13 42.422 47.692 52.963 6.254 8.7215 14.397 2.4189 -13.87 -7.192 -0.511 5.4022 7.1036 10.375 3.1921 0 pups pups 1 Diff (1-2) T-Tests Variable Method Variances pups pups Pooled Satterthwaite Equal Unequal DF t Value Pr > |t| 19 15.1 -2.25 -2.78 0.0363 0.0138 Equality of Variances Variable Method pups Folded F Num DF Den DF F Value Pr > F 12 7 11.57 0.0034 ***************************************************************************** One can use PROC MIXED to get the separate variances confidence interval. “iup” is the grouping variable and “pups” the criterion variable. PROC MIXED; CLASS iup; MODEL pups = iup / DDFM = SATTERTH; REPEATED / GROUP = iup; LSMEANS iup / PDIFF CL; RUN; -----------------------------------------------------------------------------------------------The Mixed Procedure Type 3 Tests of Fixed Effects Effect Num DF iup 1 Den DF 15.1 F Value 7.75 Pr > F 0.0138 Differences of Least Squares Means Effect group _group iup 0 1 Estimate Standard Error DF t Value Pr > |t| Alpha Lower Upper -7.1923 2.5831 15.1 -2.78 0.0138 0.05 -12.6951 -1.6896 Verification that SPSS and SAS correctly computed the separate variances error term: 2.5635 2 8.7215 2 2.5831 8 13 Karl L. Wuensch Psychology, East Carolina University 26. November 2008