I have two intervention groups with uneven subjects. When I run a two-way RM ANOVA it takes the means of the smaller group thus I presume the ANOVA results are incorrect. Is there a way of accounting for the different, but actual, sample sizes in SPSS?
No, an ANOVA is appropriate. You just need to run it as a 2X2 ANOVA with both repeated measures and between subjects measures.
In SPSS, the easiest way to do it is:
1) Choose Analyze --> General Linear Model --> Repeated Measures
2) Add a 2-level factor, and then click "Define"
3) Add the pre-post measures in the Within-Subjects Variables box
4) Add the control / intervention measure into the Between-Subjects Measures box.
5) Add the appropriate Post-Hoc or Contrasts based on the a-priori theory of your study
6) Click OK.
Can you clarify the research design? Do you have both repeated measures AND between-subjects measures? Because if you only have repeated measures, it is impossible to use different sample sizes. The very nature of the "repeated" in repeated measures means that there must be values for each subject in both groups. any subjects that only have a value in one group but not the other gets thrown out as invalid data.
Hi John! Thank you very much for your response!
I have two different test groups (control and intervention), and I took the same measurements from the subjects in both test groups pre and post-intervention. However, I have one less subject in the control group.
I have previously used a paired sample t-test but I would like to use an ANOVA which I thought was a two-way repeated measures ANOVA. Maybe I'm using the wrong statistical test?
Thanks.
No, an ANOVA is appropriate. You just need to run it as a 2X2 ANOVA with both repeated measures and between subjects measures.
In SPSS, the easiest way to do it is:
1) Choose Analyze --> General Linear Model --> Repeated Measures
2) Add a 2-level factor, and then click "Define"
3) Add the pre-post measures in the Within-Subjects Variables box
4) Add the control / intervention measure into the Between-Subjects Measures box.
5) Add the appropriate Post-Hoc or Contrasts based on the a-priori theory of your study
6) Click OK.
Hi John.
Thank you so much! That's brilliant!! Really appreciate it.
One final question, should my data in SPSS therefore be still set up as if it was a t-test (pre control, post control, pre intervention etc.) in separate columns or should it be the parameter values be in one column and the other columns coded for time (1.0 for pre, 2.0 for post) and intervention (1.0 control, 2.0 intervention)? Hopefully that makes sense.
Thanks.
If you're using SPSS, each subject should be a different row, and the subject's pre- and post- values should be in different columns.
Whether the residuals are studentized won't influence the ANOVA. It does however impact on the diagnostic plots that you can obtain through SPSS. It may also make a difference if you are planning to exclude data based on extreme residuals (not that I'd generally recommend that). Generally studentized residuals are thought to be more appropriate - though it won't often make much difference.
If you genuinely have unbalanced data in a repeated measures design (usually through missing data) I'd recommend a multilevel (also known as linear mixed) model
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS