Independent t test - small sample size - use all data points?

BOAZ

New Member
#1
Hi there,

I am running a small study in which both treatment (n = 7) and control groups (n= 6) have undertaken the same 48 item rating scale task (scale = 1 - 10) pre- and post-treatment period. We are running an independent t test to see whether the mean difference score (post test score - pre test score) for the treatment group were significanly larger than for the control group.

Rather than running the t-test on the mean difference scores for each particpant (this provides only 7 vs. 6 data points), I am considering whether to run a t-test on all all differnce scores collected. This would give me 336 (7 x 48) for the treatment group and 288 (6 x 48) data points for the respective group. I could then run an analysis that is more robust to violations of t-test assumptions (non-normality, unequal variances, unequal group sizes etc), and of course run robust checks for these vilations.

However, I am unsure if this is a correct way to go about the analysis, given that in SPSS I will have more than one score per particpant in the same column. So my question is: Can I run the analsysis on all difference scores, rather than the mean difference scores for each participant?

Any advice would be much appreciated.
 

Karabiner

TS Contributor
#2
It is not possible to treat several measurements
taken from the same subject as independent
observations. Otherwise you would not have to
bother to collect data from 6+7 people, just 1
versus 1 would have been sufficient, if only you
had enough measurements from these two
persons. If you treat 48 measurements from the
same subjects as if it were measurements from
different subjects you just produce meaningless tests.

Kind regards

K.