significant differences between 2 groups

#1
Hi,

I have been trying to find out how I can most significantly analyse the differences between two groups in my survey (like male and female) on a range of about 32 questions that are all ordinal (Likert scale). I would like to find the best way to analyze the 32 questions and find out which of these have a significant difference between male and female.
I have been reading so much that I have become confused. Could I use Wilk's Lambda on the data, usually used for discriminant analysis? Kolmogorov-Smirnov maybe or even Chi-Square? Does anyone have an impupt on the best measure to use?

Thanks for your help guys.

Leila
 

JohnM

TS Contributor
#2
Actually, you could treat these as interval scales without much concern for messing up your analysis or conclusions.

At first, keep it simple and do a graphical comparison of the means of each gender for each question - the overall pattern of results, i.e. which questions tend to be higher for males or females, may prove interesting...

Maybe then follow up with t-tests on each question to see which, if any, of the graphical comparisons are statistically significant. The reason I suggest a t-test relates to my first paragraph, and the fact that I believe that the nonparametric tests will not be powerful enough to see any differences....

Finally, this may be a good candidate for MANOVA, which will tell you what linear combination of survey questions best discriminates between males and females.