I am having trouble determining which test to run against our customer survey results. We had them answer about 20 questions having to do with our services, indicating on a 1 to 5 point scale (re-designated as scores of 0,25,50,75, or 100 on a Net Satisfaction Index), their opinion given a range of 5 responses from "Very Dissatisfied" to "Very Satisfied."

So, what we have is a table with columns like "Innovation," "Commitment to Customer," and "Overall Satisfaction." The rows are individual cases. Each customer has graded each question, resulting in scores of 0,25,50,75 or 100 for each. What I want to do is find which of the categories (questions) most drives the "Overall Satisfaction" category, i.e. which question's answers best predict each customer's answers to the main one, the "Overall Satisfaction" question.

I've been looking at correlations using Spearman's Rho and Kendall's tau_b, but don't know for sure which is better, if either, nor how precisely to explain to management the meaning of the numbers.

Am I running the right tests? How, in layman's terms should I describe the results?

Thanking you in advance...

Ed

From www.ruf.rice.edu/~lane/hyperstat/A62436.html (http://www.ruf.rice.edu/~lane/hyperstat/A62436.html) and http://davidmlane.com/hyperstat/index.html

Also see: www.statlets.com/usermanual/glossary.htm (http://www.statlets.com/usermanual/glossary.htm)

I'm not convinced you will find a correlation. What have you found so far?

<FONT COLOR="BLUE"><BLOCKQUOTE>Am I running the right tests? How, in layman's terms should I describe the results?</BLOCKQUOTE></FONT>

Yes, you are running the right tests. But, I would suggest a t-test to confirm the relationship between the variables. I would not worry so much about correlation as about a direct relationship between your variables. Just a thought.

Describe the results in terms that your audience will understand.

Regards,

Don