I have a study and I'm not sure how to do a correlational analysis on the survey data that I have from it. There are two independent variables and a dependent variable. Each variable is being measured by a Likert scale survey. The scales do not have the same number of ratings, as one is a 1-5 Likert scale, one is a 1-6 Likert scale, and one is a 1-7 Likert scale. These three surveys were incorporated into a single survey sent out via Survey Monkey with demographic questions. Can you even do a correlation analysis on this? I would assume Spearman’s correlation, but I am hung up on the fact that I have three surveys with different scale lengths. What is the best way to compare the data between these scales?
Correlation measures are invariant to linear transformations. This is true for the Pearson but also for Spearman correlations. In other words you can change the scale of one variable and the correlation stays the same, as it is the case with your three different scale lengths. So, apply Spearman, and do not worry about different scales.
You can correlate miles with kilometres, so you can also correlate a 5-point scale with a 7-point scale, and with a 6-point scale.
More important is that a Likert scale - the sum or average of several Likert scale items (questions) - is not an interval scale. It's an ordinal rating scale. We social-science folk argue that, with multiple items (to smooth out random error within respondents) and a reasonable sample size (to smooth out random variation among respondents) then the Likert scale "approaches" interval properties. So, to be technically correct, you would use Spearman's Rho rather than Pearson's r to calculate a correlation coefficient. (but you'll probably find that they're about the same.)
- Badges
- Science topic
- Similar topics
- Statistics
- Survey