In my logistic multilevel modelling there are four variable of interest( measured in likeert scale). I wish to handle the model in two ways. Firstly making an index out of the four variables and in a second model I wish to see the effect of individual variable.

In the first instance I wish to take the normalized score of each variable and form an index. Is it a right idea. Or I need to form an index out of original data without normalization? can I divide the index into two parts on the basis of mean or median? If I donot divide the index, can I use c. command ( telling stata to treat the index as continuous variable) for the normalized index while taking interaction with any other categorical variable like age-group or gedner?

The next challenge is what should I do to check individual effect of four variables as I have mentioned? I tried to create4- dummies for each variable.Now we have four dummies for each variables. Next can I take only the 4th dummy of each varibles to generate a new interaction variable ( that will show how highest scales(_d4) of likert scale would affect the dpendent variable if interacted together? My likeert scale is : 1= strongly dislike, 2= dislike, 3= like, 4= strongly like). If I get any suggestion from anyone , I will be highly obloiged.

Regards

More Prafulla Nath's questions See All
Similar questions and discussions