Help with figuring out appropriate statistical procedure for this analysis

[mitrakos]

Hi Everyone,

I have been tasked with looking at the results of a survey for this year and seeing what characteristics are associated with the lowest scores of the survey. These characteristics include multiple demographic variables, as well as other characteristics specific to the survey itself. We are looking at three specific questions.

To expand on low scorers, this survey is scored via top-box scoring. What that means is that for each question there is one (or multiple for rating scale questions) response that is dignified as the most optimal answer for the participant to answer with. Therefore, for seeing what characteristics are associated with low scoring, we are looking at characteristics that have the lowest percentage of top-box scoring for that item. This survey is a feedback survey, which is why there are optimal responses for participants to have.

I am thinking this just seems like a simple item analysis that I can conduct, but I just want to run it by this community because it has helped me so much in the past in case I am under (or over) thinking it. Thank you so much!!

[ballardw · Accepted solution]

If you have a numeric score calculated for each observation I might start with something like:

proc freq data=have;
   tables score*var1*var2*var3 /list missing;
run;

By default the score variable would appear in increasing order so the lower scores, which you state are of interest would appear first. Then the levels of the other three variables of interest associated with each score.

This is exploratory in nature to get a feel for the data. If your set is "small" it may provide everything you need.

It might point to 2 levels of one of the Vars as similar (or 6 or whatever).

[mitrakos]

Thank you so much!