TTest with small unequal sample size

[nodak]

I am trying to compare means of two groups.  The sample size for each group is small (between 9 and 15) and the sample size for each group is not the same.  I have been using proc ttest and looking at the satterthwaite correction p value. The variances are equal (normality is a little off), but I read somewhere that satterthwaite corrects for both variance issues as well as unequal sample sizes. I have also been using proc npar1way and looking at the Wilcoxon results, which I think weights by sample size as the default.  Am I on the right track?  Any advice is appreciated.  Thanks. 

[ballardw]

With samples that small I personally would lean heavily on the nonparametric approaches.

[Reeza]

Besides non-parametric, I would also be relying heavily on knowledge of the data and make sure to include confidence intervals.