Scenario - There is an IV and DV. IV is measured in 5 point likert scale questions and DV is measured in 7 point likert scale questions.
Doubts -
01. Can we run a test like regression analysis directly irrespective of differences in measures?
02. if NOT, what are the transformation techniques available to transform data into same scale?
I don't see any reason why you couldn't run a regression on these two variables. In regression, one variable could be frogs per pine tree. And one could be temperature of my coffee.... Am I missing something ?
If you have only one item for each of your measures, then you have two ordinal measures and you should not use a procedure such as regression that assumes you have interval-level data. However, if you actually have multi-item measures that you use to form a scale, then you can create interval-variables and use correlation and regression.
Different scales are not a problem here (and if they were, you could simply z-score both variables, among other possible transformations). I assume you are talking about two multi-item scales, not just one Likert item for the DV and another Likert item for the IV.
Use standardized coefficients when interpreting the results. Not the just coefficients. :)
You can simply convert responses into percentages and run the model.
X - X(min)/(X(max) - X(min)
For example, if the response is 3 on a 5-point Likert scale, the percentage is 50%. (3-1/5-1 = 50%)
No problem Harshaka, you can perform the regression without any issues. First, the data should meet the assumptions (Linearity, Normality etc) of the regression estimation technique (OLS, MLE). Second, interpret the standardized coefficients. Method bias/variance literature emphasis to change the scales of different constructs to avoid such issues at minimal (despite recommendations exist on more stringent techniques. Some of the work by Fornell emphasis on the method bias issue.
- Badges
- Science topic
- Similar topics
- Mathematics