The second and third research question in a study that I sent to a journal asked whether the interaction of the intervention with gender (two values) and ability (three values) affected significantly the outcome variable. The research involved one experimental group and two control groups. To answer the second and third research questions, I ran two way ANCOVA. The dependent variable was the outcome variable. The two independent variables, in this case, were the class type (three classes, one experimental and two control) and student’s gender (or ability). The covariate is the value of the outcome variable at the beginning of experiment.
One reviewer suggested using MANOVA to study the interaction instead of ANCOVA. Ca n you please give me your opinions regarding what statistical text is better? Would the text differ if I asked about the influence of the intervention on the two genders?
Hello Wajeeh,
So, was it the case that RQ2 and RQ3 used the same IVs (gender, intervention) but evaluated different DVs?
If so, and if the DVs are logically / theoretically related, a man(c)ova might be useful. If they are not, then the separate an(c)ova analyses would suffice.
However, if you have just one DV, and the reviewer was suggesting using the baseline DV as a second DV instead of as a covariate, I don't see how that would improve the quality, suitability, in interpretability of the analysis.
Good luck with your work!
Dear Wajeeh
MANOVA is the case when you want to measure the effects of independent variable (s) on more than one dependent variable. In your case, I see there is one dependent, two grouping independent and a covariate. So ANCOVA should be used. But first you should set the average of your covariate to ZERO (or standardize it) to reduce redundant variation. For example if the average of your DV at the beginning is 12, you should subtract 12 from all your outcome values at the beginning of the experiment. You should determine four interaction effects in the model though. gender*ability, gender*covariate, ability*covariate, gender*ability*covariate.
Hope it helps
Ehsan
Dear David and Ehsan,
First, thank you very much for your thorough answers. In fact I have four dependent variables (freedom, equality, engagement and justice) and a fifth variable consisting of the four (democracy); i.e. the four are components of the overall, which is the fifth. I understand, from David explanation, that MANCOVA fits here, taking into account the standardization that Ehsan Mentioned. I am right?
Please if you would to know the difference here :
https://www.youtube.com/watch?v=Q116ZnLy5uA&t=100s
ANOVA, ANCOVA, MANOVA and MANCOVA: Understand the difference
Hello Wajeeh,
If you're analyzing the five DVs as a set, there might be a problem, since the fifth apparently is a linear composite of the other four. Most statistics software packages with man(c)ova would note this as a problem. SPSS, for example, would just drop one of the DVs and not mention it!
Good luck with your work!
I really thank you Khaled and David. I really benefited from all the discussion here.
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS