Hello,

I think you mean "ordinal logistic regression", because the scale measure is "low, medium and high". So you can deal firstly with : univariat, bivariat ...and be carfule for the multicollinearity problem (as you cited that you have 20 IVs).

after such steps, you select the pertinent IVs to be included in the multivarite step.

see this article:

https://link.springer.com/chapter/10.1007/978-3-319-19425-7_13

Chellai F

You might also want to run a multiple regression on your continuous outcome measures. I would suggest to try an all possible subsets regression.

See, e.g.: https://statisticsbyjim.com/regression/guide-stepwise-best-subsets-regression/

or:

http://www.maths.qmul.ac.uk/~bb/SM_I_2013_LecturesWeek_10.pdf

Hi Nivine,

You can and you should perform an ordinal logistic regression. I suggest the proportional odds logistic regression model.

Here you can see how it is performed in R:

https://towardsdatascience.com/implementing-and-interpreting-ordinal-logistic-regression-1ee699274cf5

Good luck with your work!

I agree with Eliana. Yet, I would like to remark that when you categorize your continuous variables, you loose information. You might first inspect the correlation matrix of the predictors. Maybe you can reduce the number by eliminating the one who highly correlate. Maybe you can use aprincipal components analysis to help you find relevant dimensions. This could help you choosing a reduced model of predictors for the multiple regression.

Thank you all for your suggestions, I highly appreciate it.