MANCOVA help!

#1
Hello,

Can anybody help me out? I have had my head in books for ages now and I feel like I am going a little mad! They are either too basic or too complex to answer my questions.

As part of my MSc I have performed a cross-sectional survey and I am looking to analyse the following variables using MANCOVA through SPSS:

Predictor (IV): Supplement user status (Supplement taker vs non-supplement taker)
DVs: BMI, alcohol score, exercise score, no. of fruit/veg, and smoking status (which given it is categorical I will analyse separately).

I took measurement of the following as possible covariates: age, gender, number of supplements taken.

1) How do I know if the covariates are suitable? I have performed separate ANOVAs with each possible covariate as the DV and my original IV as the fixed factor (supplement status). I would be looking for a non-sig difference to prove independence and hence use as a covariate? Is what I am doing right?

2) Can I use gender as a covariate given it is categorical? If not how would I go about controlling for it?

Thanks. :)
 

Karabiner

TS Contributor
#2
DVs: BMI, alcohol score, exercise score, no. of fruit/veg, and smoking status (which given it is categorical I will analyse separately).
Why MANOVA? Do really these variables jointly represent a hypothetical construct
which you want to study?
1) How do I know if the covariates are suitable? I have performed separate ANOVAs with each possible covariate as the DV and my original IV as the fixed factor (supplement status). I would be looking for a non-sig difference to prove independence and hence use as a covariate? Is what I am doing right?
i must admit that I not quite understand your rationale.
Why don't you just put them in because of theoretical considerations?
2) Can I use gender as a covariate given it is categorical?
You can use it e.g. as a dummy-coded covariate.

With kind regards

K.
 
#3
Thanks for your reply!

The DVs represent health behaviours as a whole- I am basically looking to consider whether individuals who consume dietary supplements are more/less healthy than those who don't.

In terms of adding covariates for theoretical reasons, there is limited theory in this area as yet. It just seemed logical that for instance gender/age may influence exercise and BMI etc.
The rationale I explained is the procedure stated in stats books for an ANCOVA. I just can't find whether that necessarily applies/follows in the same way to MANCOVA. Like normal variables, I was under the impression covariates had to be checked for assumptions as well- so really that is what I am confused about how to do.


All the best,
Emily.