Hi,
I have a simple question regarding the interpretation of MANCOVA in SPSS. I have found no statistical significance for the overall effect of group according to the multivariate test, but the individual between-subjects effect of the fixed variable is significant on one dependent variable.
As I understood, normally one should just report the non-significant p-value and stop there, but does the between-subjects effects test still tell you more? Or is the result of between-subjects effects not valid, since the multivariate test is not significant?
Assumptions have been checked and nothing violates MANCOVA.
Design: pretest, posttest control group, experimental group study with 3 dependent variables, a covariate and group as fixed factor.
Thank you so much in advance.
Hello Emilia Kumpulainen
Can you put SPSS software output here?
I think it is good to see it before any answer.
I would suggest just running another test (ANOVA, t-test, etc.) for the fixed variable.
Hi Abolfazl Ghoodjani surely!
Please find the SPSS output attached below. Box's M and Levene's tests are not significant.
Thank you in advance!
Emilia Kumpulainen Seeing the pictures, I realized the issue.
Assuming your analysis is correct (although I still do not understand why you did the multivariate. I think three separate ANCOVA are enough. Unless you have a specific idea of the multivariate.)
I think, contrary to multivariate (Sig = 0.228), you have a significant in pred12 (sig = 0.047).
To understand why this happens, look at the number of degrees of freedom and the value of the F statistic in your tables.
Abolfazl Ghoodjani thank you so much, this was helpful.
I chose MANCOVA, since I wanted to see if by controlling the pre-test result the group has a significant effect on the combination of all 3 dependent variables. (there is not). If I understood correctly, ANOVA/ANCOVA doesn't examine that, but can be only used for when there is one dependent variable.
In this case by running 3 separate ANCOVAs I should see the effect/no effect of the group on the dependent variable, right? And it would be appropriate to interpret that?
I run the ANCOVAs now, and get more or less the same result for the group's effect on productivity (prod12) as with the between-subjects effects of MANCOVA suggested: p=0.048 (in ANCOVA) and p=0.047 (in MANCOVA). Is it still better to run and report the ANCOVA results for the independent effects, after stating that the groups overall effect on the combination of variables is not significant?
Thank you so much. I really appreciate the help and guidance, since I am not very experienced with either SPSS or statistics in general... I attached some pictures of ANCOVA output again for support.
Emilia Kumpulainen It is clear that ANCOVA has no superiority over MANCOVA, and only the researcher's idea and opinion on how to analyze the data is important. If you want to analyze the dependents as multivariate, then MANCOVA is better.
- Badges
- Science topic
- Similar topics
- Biological Science
- Histology
- Fixatives