# Likert scale questionnaire and logistic regression

I have three independent variables , and one moderating variable. I'm studying the affect of these variables on the decision to adopt electronic commerce.

I have a questionnaire, based on this questionnaire which is on a 7-point likert scale (from totally disagree which is coded 1 to totally agree which is coded 7).

I have gathered data on on my three independent variables. Now I want to analyze the model.

Question is should I calculate mean for every IVs and then do the logistic regression?

Regards

-
 What's your question? (The last line appears to be a little garbled.) – whuber♦ Aug 1 '11 at 18:49

If you can assume that the responses are continous, just run a linear regression. However, It will not be a good aproximation if, for instance, people considered 7 and 6 as similar, or 1 and 2 as similar.

If you can't assume they are continous, you can run an ordinal logistic regression directly with your data, or you can agregate responses in fewer categories (in the limit, you agregate in only 2 categories and ran a logistica regression).

What makes more sense depends of your problem and data. Visual inspection of the relationsfip of response and IVs can help you to decide what to do.

-
 Hello, thanks for your answer. The point is not which analysis to perform. Indeed from the paper I read, I know that I must use the losgistic regression, non the normal regression. To answer to your quesstion, yes 6 and 7 are considered as similar. – Luca Sep 11 '11 at 10:19

There is no need to do recentering. The linear part of the GLM can simply include an intercept term (e.g. b0 + b1*x1 + b2*x2 + b3*x3), and this will be estimated by standard statistical software.

-
 Hello, thanks for your answer. Could you please be so kind to make an example in R with the glm function you mentioned? So I will understand better. – Luca Sep 11 '11 at 10:20