If I understand you correctly you would have better luck treating this as a multivariate regression

See the classic Johnson and Wichern book on multivariate statistics. David Booth

https://statistics.laerd.com/spss-tutorials/one-way-mancova-using-spss-statistics.php

I guess this could help you.

You have several options and it depends on your emphasize which parameters are important to know/show. Basically you have two sets of variables, 4DVs and 1IV (special case that this is only one variable). Therefore you can do either a MANCOVA without fixed factors and only the IV as covariable, or a Canonical Correlation (with special case that the IV side has only one variable) or you switch the logic and and use multiple regression and treat the DVs as predictors and the IV as criterion (the former two could be considered as a form of multivariate regression as David mentioned). All will have the same omnibus test and are therefore the same basic model. But the difference is how the "post hoc" tests are calculated and what SPSS will show you. There is now no inherit "correct" approach, but it depends on what you want to show.

Am I the only one who thinks it's not worth answering questions from a fake profile ?

Also it seems that you can find the answer to your question doing a google search, which is against the guidelines of research gate

Stefano Nembrini I didnt even notice the apparently fake profile, typically I dont care who asks a question. But you are generally right.

Thank you for the response, Rainer.

For my analysis the IVs are 6 subscales of a validated scale and my hypothesis was initially to show that these predicted 3 of the DVs, which in turn predicted 4th DV.

So, I've done the MANCOVA with only covariates and I supposed the post-hoc tests would be the same type of posthoc tests for any MANCOVA analysis?? The part where i get confused is when post hoc tests start referring to differences between groups

Stefano Nembrini Yes, you are right this is a "fake" profile (if you assume anonymous and fake are synonyms) but I am nevetheless still a real person. I just felt embarassed for someone in my position, to publicly post a statistics question that I should probably know the answer to already... I had tried searching this question on google and for MANCOVA analysis, etc. But every example I found included 2 or more groups.

This does more sound like an structural equation modeling approach than a regression/correlation approach, since you have several variables that are sequentially dependent on each other.

If I got you correctly, you have 6 (continous?) IVs, which will predict 3 DVs, which in turn predict the 4th DV. Without knowing more about the hypotheses and the intended structure, it is hard to say if this should be analyzed as a path analysis, with a mediation part (the 3 DVs in the middle would be parallel mediators), i.e. a formative measurement model or as a confirmatory factor analysis with primary and secondary factors (i.e. reflective measurement model). With SEM, you can also decide which IV should predict/load on which other variables individually. You should also consider if they are recursive or non-recursive to each other.

I think SEM would be a much better and flexible approach, but this would be beyond the scope of this forum. Please consult a local statistician for help, and have a look at relevant textbooks on that topic, such analyses can become quite complex.

I understand your situation

Sam Unknown this being said, not using your real name is against the guidelines

https://www.researchgate.net/community-guidelines