2

I've run a 2 (treatment and no treatment) x 2 testing occasions (pre and post) Repeated Measures ANOVA. In addition, I entered several tests given at both time points, so under the Repeated Measures Define Factor(s) window option for Measure Name, I've entered 5 different tests used.

In the output, I am trying to determine which table indicates the main effect across all tests. What I am seeing instead is a Tests of Within-Subjects Effects Multivariate table (that I am ignoring) and a Univariate table that lists session, session*group, and error(session) effects by test.

To report a main effect to address the question of whether any tests differed between the groups from pre to post, where would I find that information?

jonsca
  • 1,772
redracer
  • 67

1 Answers1

4

SPSS usually provides univariate tests of such a main effect on each variable all the way down in the output (“Tests of Between-Subjects Effects”), even for doubly multivariate designs. So, barring any particular problem in the way you specified the model, they should be there. There are also multivariate tests of between-subject factors (why are you ignoring them?)

Furthermore, from your description at the end, I am not sure that you should be looking at a main effect. It could be that the interaction in fact addresses your research question, see my answer to Experiment with two groups, pre- and post- treatment assessments.

There is also a lot of relevant material on this site that you might want to read, in particular Best practice when analysing pre-post treatment-control designs

Community
  • 1
Gala
  • 8,501
  • Thanks for the feedback, but even in the Tests of Between-Subjects Effects, the results are listed for intercept, group, and error by the measures I entered, not as a composite for what I expect to see for an overall main effect. The only output that does not report stats by the independent measures that I used are 2 Multivariate tables. One is listed after the Within- and Between-Subjects Factors, and the other is nested under Tests of Within-Subjects Effects. I'd like to know how to find the correct stats to report if there are significant main effects or interactions for group or time. – redracer Jun 15 '13 at 17:26
  • Well, conceptually this “composite” test is a multivariate test. You do have several variables after all so you have no choice here. Whether this is the “correct” stat will depend on what you want to know and what you are willing to assume. Multivariate and univariate tests have slightly different assumptions. – Gala Jun 15 '13 at 17:39
  • @redracer Looking at your past questions, I see that you never accepted an answer and tend to resist the advice you are given. What you are asking (a univariate test of an inherently multivariate hypothesis) makes little sense to me. Again, why don't you want to consider multivariate tests? – Gala Jun 16 '13 at 08:34
  • I've been researching more, thanks. I am deciding on: 1) whether it is better to run separate ANOVAs for each test; and if not, 2) how to write up the multivariate and univariate results. Here is a condensed example: "a 2 x 2 repeated-measures ANOVA was conducted...using multivariate and univariate statistics. Intervention effects differed depending on group assignment (F(5,37) = 6.076, p < .005). [<--Multivariate result] Interactions were significant for test 1-test4 (p<.05). The interaction was not significant for test 5, with both groups declining from pre to post (p=.6).[<--Univariate]" – redracer Jun 16 '13 at 14:09