Dear Allstaters,
I am analysing a survey with 50 questions under 5 categories (such as opinions relating to Organisation, Corporate Culture, Supervisory Practices etc) with the responses Very True, True, Sometimes True, Seldom True, Never True, and Don’t Know.
At the start of the survey, various demographic questions (i.e. age, gender, years at company) were asked. Now, as well as the usual frequency distributions of the 50 questions by sub-population, I was thinking of performing the following regression analyses:
- Transform the response categories into an ordinal response (i.e. Very True=1, … Never True=5) but set Don’t Know to missing.
- Pool the responses under each category (i.e. to Organisation, Corporate Culture, Supervisory Practices etc) and calculate the mean response for each subject.
- Analyse a GLM with mean response for each category as the dependent variable with age, gender etc added as the independent variables.
- The idea would be to investigate how demographics affect opinion relating to the 5 categories.
The one concern that I have with the analysis proposed above is the implications of setting questions answered Don’t Know to missing.
I would appreciate if you could send me a direct response relating to your ideas relating to the analysis above, suitable literature to read, and alternate methods of analyses. I will then send out a pooled response to my query later.
Thank you.
---------------------------------
Do You Yahoo!?
Yahoo! Finance - Get real-time stock quotes
|