topic Re: Comparison vs a baseline in Statistical Procedures

Comparison vs a baseline

deleted_user — Mon, 09 Nov 2009 15:11:15 GMT

Hi all:

For a number of variables, I need to compare the mean of 4 individuals vs the baseline of those same individuals. I'll have several occasions where measurements were taken, so the analysis would be applied to each occasion.

The idea is to see whether there is a change of the mean (at each occasion) with respect the baseline.

I guess a t-test is not the option. I really appreciate your help.

Kindest regards.

Re: Comparison vs a baseline

Paige — Mon, 09 Nov 2009 16:02:44 GMT

PROC TTEST with the PAIRED statement should work. If you want to get more fancy, you could used PROC MIXED with the REPEATED statement.

Re: Comparison vs a baseline

deleted_user — Mon, 09 Nov 2009 20:16:09 GMT

Thanks, Paige.

I also found a book that would help me (Common statistical methods for clinical research). It seems I have to verify normality before deciding to use a one-sample t-test. If normality is not the case, I have to go with Wilcoxon signed-rank test.

My problem is that there are very few individuals, so proc mixed and repeated measures may find convergence problems (?).

Regards

Re: Comparison vs a baseline

Paige — Tue, 10 Nov 2009 13:10:53 GMT

Not sure you can "verify" normality with 4 individuals. With 4 individuals, I doubt that whatever 4 numbers you have would fail a normality test. Just seems unlikely to me with real data. I would do the t-test, myself.

If you are worried about convergence problems in PROC MIXED because of 4 individuals (which is something I don't normally worry about), you should probably also be worried about the validity of Wilcoxon on 4 individuals; and also worry about the power of the test being very low.