How to analyze data with more than one associated categorical dependent variables?

#1
I have some dependent variables related to the growth of a company having categories like (e.g. for variables indicating net profit, financial turnover etc.)

(1) decreasing,

(2) stable,

(3) increasing and

(4) rapidly increasing.

Now these variables are associated and ordinal in nature. I want to see the effect of a set of co-variates on these dependent variables. I understand that separate logistic regression will result in wrong formula of the variance and thus will produce wrong p-values. But what kind of regression should I run with more than one associated dependent categorical variables? Should that be a GEE model? Or anything else like multivariate regression (GLM)?

Can SPSS or R perform this?

Besides, I would like to have your suggestion regarding whether I should rather consider these variables as nominal and take "stable" as the reference category so that interpretations are more meaningful?

Kind regards.
 

Karabiner

TS Contributor
#2
Re: How to analyze data with more than one associated categorical dependent variables

SPSS (and then of course R, I suppose) can perform something that is called [multiple] ordinal regression.

With kind regards

K.
 
#3
Re: How to analyze data with more than one associated categorical dependent variables

Dear Karabiner, could you make it a little more clear? Can I use, say, 3 dependent variables simultaneously there? I know that multiple ordinal regression is an ordinal regression that have multiple independent variables. But in this case I have multiple dependent variables which are further associated. Could you please explain me a little broadly? That will be a great help indeed.
 

Karabiner

TS Contributor
#4
Re: How to analyze data with more than one associated categorical dependent variables

Oops, sorry, I mis-read your post; thought you had just 1 ordinal dependent variable.

I was once told that GEE can handle repeated logistic regressions, so your idea might
perhaps lead you in the right direction.

With kind regards

K.
 

noetsi

Fortran must die
#5
Re: How to analyze data with more than one associated categorical dependent variables

I don't know for sure if any of these methods will do what you want, but you might look at:

MANOVA (which is a form of ANOVA that deals with multiple dependent variables)
Canonical Correlation (a method that does not get talked about much)
Structural Equation Models which can handle multiple dependent variables although its a good deal more complex than regression IMHO and requires speical software such as AMOS or Mplus. SEM uses regression and I believe (although I Did not work with this form) can use ordinal variables in certain methods.