1
$\begingroup$

I would like to analyze data I collected through a survey questionnaire and I'm stuck on the type of regression model to use.

The task is to do a regression analysis where my dependent variable is a Likert scale, so it is made by the mean I computed of three items measured through a 5-point scale. As my Dependent variable is not a single Likert type item, but it is multi-item based, I thought it was fine to use the linear regression model and not the ordinal regression.

Any information on the same is appreciated.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

I would recommended not averaging over items and instead using an oridnal model with random effects. There are a bunch of problems both averaging over items and using metric models on ordinal data (e.g., Liddell & Kruschke, 2018).

You can use the 'ordinal' package in R (Christensen, 2015). It can handle random effects for items and participants so that you don't have problems with pseudo-replication.

If you would like a tutorial, this one was useful for me: Bürkner, P. C., & Vuorre, M. (2019). Ordinal regression models in psychology: A tutorial. Advances in Methods and Practices in Psychological Science, 2(1), 77-101.

Other things I cited:

Christensen, R. H. B. (2015). Analysis of ordinal data with cumulative link models—estimation with the R-package ordinal. R-package version, 28.

Liddell, T. M., & Kruschke, J. K. (2018). Analyzing ordinal data with metric models: What could possibly go wrong?. Journal of Experimental Social Psychology, 79, 328-348.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you for your answer! The fact is that the 3 items I used in my survey were taken from previously validated studies that averaged them to get the variable of interest and I did in fact the same, along with factor analysis and Cronbach to be sure of construct validity and reliability. My issue is that having a multi-item based likert type DV, I thought this type of variable was an approximation of continuos data and so linear regression can be fine... in fact on the contrary how could I run an ordinal regression if my DV is multi-item based? $\endgroup$
    – Jack
    Commented Dec 16, 2020 at 11:04
  • $\begingroup$ If you only have data that is averaged over multiple items I'm not sure what you could do at this point. If it's standard in your field you may not have pushback from reviewers, but as discussed in Bürkner & Vuorre tutorial, averaging over items doesn't help the issue at all. $\endgroup$
    – sjp
    Commented Dec 17, 2020 at 0:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.