Paired Samples T test or Repeated Measures ANOVA?


New Member
Hi everyone,

I am on a tight deadline and could use some input on my proposed analyses for my project.

PROJECT: I am administering a novel treatment to participants and seeing if they lose weight at the end of 6 months of treatment. There is no control group (there are rationale for this, just not relevant to my question). I will be collecting weight data as well as administering a battery of measures looking at behavioral and psychological outcomes. Participants will be assessed at pre-treament and post-treatment.

GOAL: Determine if there are significant changes in weight and other DV variables (eating habits, physical activity, etc.) from pre-treatment to post-treatment

QUESTION: Should I do paired samples t-tests (i.e. pre- post- comparisons for each of the DVs) to evaluate whether or not my treatment is effective
OR should I do a repeated measures ANOVA or a one-way ANOVA?

Here's my thought process: I understand that a basic one-way ANOVA (not repeated measures) requires independence of samples, which we don't have since we're looking at pre- post- data within the same participants. Further, my understanding is that repeated measures one-way ANOVA requires more than two groups--we only have two groups since it is pre- post-data only so a repeated measures ANOVA would not be appropriate, correct? Would it not be appropriate to run paired t-tests on the pre-post data? Also, since we're running t-tests for separate outcome variables, Family Wise Error Rate should not be a problem, right?

Any assistance would be greatly appreciated. Thanks for having a look!
Repeated measures is >=2 not >2. So either will work and with 2 groups the results are equivalent. The F you calculate should equal the paired t squared. Or in other words, t = sqrt(F). It shouldn't matter which you use except that the t test will be easier to calculate (theoretically).

One-way between subjects ANOVA is inappropriate. I don't think family wise should be a problem, but: Because you mention multiple dependent variables, if you are interested in their relationships, you might want to do a multivariate test if that is of interest.