Hello,
I have an experimental research design with 1IV, 2DV and 3 covariates for my master thesis. The normality and homogeneity of variance assumptions are violated, so I cannot run a MANCOVA to see the interaction effects.
Is there any alternative analysis that would be suited in this case?
Two separate ANCOVAs?
Or what would be the best way to proceed?
I appreciate your answers!
Hello Diana,
The key point is, what specific research question/s do you seek to address by your analysis? This, coupled with the nature of the variables involved, should guide you as to whether it is most appropriate to evaluate your data set univariately or multivariately, and by what method.
You probably know that the normality assumption applies to the distribution of model residuals (e.g., difference between actual DV score and DV score as estimated by final model), so I presume you're checking appropriately this assumption from the outset.
If you're convinced that your data do not meet the usual parametric assumptions then your options are:
1. Use an exact/permutation test. Have a look at the free R library, MANOVApermtest: https://rdrr.io/cran/MAINT.Data/man/MANOVAPermTest.html)
2. Use a resampling/bootstrap test (there are far more of these available for various statistics packages than suitable versions of #1). Here's a link that may be helpful: Article Parametric and nonparametric bootstrap methods for general MANOVA
3. Have a look at "hybrid" versions of parametric tests, as proposed by Conover and Iman (the easiest one to implement), and those by Puri and Sen (only slightly more difficult to implement).
4. Look for a non-parametric alternative.
5. Consider robust regression models (for multivariate: https://www.jstor.org/stable/25470844; there are plenty of univariate examples available).
Good luck with your work.
Running a MANCOVA may not be appropriate if the normality and homogeneity of variance assumptions are violated. However, there are some alternative analyses that you can consider:
It is important to note that each of these alternatives has its advantages and limitations, and the choice of which analysis to use will depend on the specific research question and the characteristics of your data. You may want to consult with your advisor or a statistical consultant to help you choose the most appropriate analysis for your study.
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy
- Crop Science