Beginner needing basic advice - Ordinal Data correlations - how??

#1
Hi all,

I'm very new to statistics and was hoping for some easy to digest infomation regarding my (probably easy) problem. I have a set of variables that have all been measured on a 5 point likert scale, and want to test if there is any corrolation between them (as one increases, does the other).

However, I am also wanting to account for some control variables (gender and one other yes/no question).

So basically, I have multiple sets of ordinal data I want to see if there is a correlation between while accounting for 2 control variables.

How would I go about doing this?

Thanks in advance
 

katxt

New Member
#2
If the data were normal, then you could check if two variables were significantly correlated, while allowing for others, by using a general linear model or, equivalently in this case, multiple regression with the control variables coded 0/1. Likert data isn't actually normal, but it is usually reasonably well behaved, so this approach should work so long as you have a reasonable amount of data. Otherwise, you could do the regression using some variety randomization test.
A perhaps more important consideration is the problem of multiple testing. If you have say 5 variables, by the time you have checked each with each, you have done 5x4/2 = 10 tests. You are quite likely to get a significant correlation at the p = 0.05 level even if there is no connection anywhere. You will need to decide what to do about this problem.