Hi,


I need to perform regression on customer satisfaction data vs other drivers. These are following data points which are from the customer surveys received.

10 - best
5 - acceptable
1 - bad

Overall Satisfaction; Values range from 1 to 10


Secondary drivers (Values range from 1 to 10)
Clarity of instructions
Understanding issue
Technical knowledge
Taking ownership
Phone agent professional handling
Timely manner
Time to resolve issue - phone

What type of regression analysis can i perform to identify the critical drivers for customer satisfaction? I am using Minitab.

Since both X & Y data are discrete variables, your appropriate analysis method would be the Chi-square.

Standard regression analysis requires both X & Y data to be continuous. If X is continuous and Y is discrete, you can use Logistic regression analysis.

Since both X & Y data are discrete variables, your appropriate analysis method would be the Chi-square.

Standard regression analysis requires both X & Y data to be continuous. If X is continuous and Y is discrete, you can use Logistic regression analysis.

Actually, you need to be careful of the Chi-square. The limitation on the Chi-square analysis is not just the they are discrete variables, but that they are Poisson variables. Counting the number of discrete Poisson events. Data scored as 1, 5, 10 would not likely be Poisson.

Standard regression analysis might work - the limitation on regression analysis is that the residuals to the fit be normally distributed. It may still work even if the data are discrete.

Personally, I'd shift the survey data to a percentage data and do Logistic regression or binomial analysis. Or, I'd make control charts if the survey data was amenable to the techniques I use on survey data at http://www.hanford.gov/rl/uploadfiles/VPP_AnaSurveyData.pdf