Hello everyone, I am using a Repeated Measures Mancova with 4 factors, each one with 8 groups. I also included 3 covariates in the analysis. I discovered there was a significant difference between at least two groups in one of my interactions. I would like to apply a Pairwise comparison to compare separately the groups of my factors, but taking into consideration the interaction. How can I do that in SPSS?
Thanks in advance.
Conceptually, no you can't do this. A pairwise comparison is a difference in levels of a factor and averages over other factors (including those involved in an interaction).
Its possible that what you want can be done with an interaction contrast. That's not trivial to do in SPSS (but is possible).
Additional points: its a bit confusing because you don't have groups in a repeated measures design so this might be mixed measures. MANCOVA can further confuse things (and generally should be avoided in my view). If you have covariates and it is a pure repeated measures design then its possble they have no impact on anything (as they just reduce the participants/subject term) - though they may do if they covariate by factor interactions are included. There used to be an issue with SPSS MANCOVA and covariates which can produce massively misleading results. I'm not sure if recent versions resolve this. There is a simple check which is to centre the covrariates (subtract the mean from each so they have a mean of zero). If the centred covariates change the interpretation for other effects then there is a problem with the original model.
Erica Castilho Grão, please post the command syntax for your model. That will make it much clearer to everyone exactly what you are doing. Thanks.
It is twenty years since I've used SPSS. However I recall when I did my masters dissertation, there was some way to assess whether it was interaction or confounding. SPSS analysis without anguish is a good book.
Deborah J Hilton, I am not familiar with the book you mentioned. Here is the web page for the SPSS 20 version, which is the most recent one I have been able to find:
- https://www.wiley.com/en-au/SPSS+Version+20+0+for+Windows:+Analysis+without+Anguish-p-9781118337769
And here are a couple observations based purely on what I see in the TOC.
Again, these suspicions are based only on what I see in the TOC. I'd be happy for any clarifications you or others familiar with the book can offer.
Thanks Bruce / Dr Weaver. I'm not sure of the version I used, my memory has failed me twenty years ago, maybe it was version 11, I'm not sure. I also found Armitage and Berry, Statistical Methods in Medical Research. Third Edition. 1994. Confounding is explained on page 268. Interaction for two factor factorial 255-7; in GLM 425-9; in multiple regression 327; and several other instances. Maybe this will assist Erica also. The book is falling apart at the binding, so all the pages are falling out. It is a bit difficult for me to comprehend also, in that it is difficult statistically, complex explanations.
- Badges
- Science topic
- Similar topics
- Correction