Friday, 21 August 2020
Question: Can I use ANOVA or linear regression if my response variable is measured on a Likert scale?
Answer: Recently, I have had several students ask me some interesting questions on planned parametric analyses of Likert scale response data. These brought up the big question of whether you can use parametric analyses, such as ANOVAs or linear regressions, on Likert data (measured on an ordinal scale) or whether you should use non-parametric analyses. The standard assumption for parametric analyses is that your response variable data are continuous (measured on either interval or ratio scales) and normally distributed. However, in some disciplines, many practitioners assert Likert scales can provide reliable outcomes using parametric tests. This is a hot topic in academic discussion, with passionate view-points on the for and against sides of the argument! Check out Gail Sullivan’s & Anthony Artino’s excellent insights in their published editorial here. Also, check out Karen Grace-Martin’s brief summary of the issue and excellent practical advice here on what to consider to make a sound decision in your own research. There are other tools though that may prove to be better! Keep in mind that generalisations to linear approaches have been developed over more recent years specifically to overcome these violations in assumptions. These extensions are called Generalized Linear Models, and JMP gives a nice introduction to them here.
Written by Chandra P. Salgado Kent Quantitative Adviser (Graduate Research School) & Associate Professor (Centre for Marine Ecosystems Research, School of Science).
Have a research methodology related question you’ve not been able to answer?
Check out available research support here or register for upcoming Quantitative Research workshops:
Online, self-paced alternatives of these workshops and an extensive range of Quantitative Research resources available via the GRS Programs BlackBoard site.
Please leave a comment about your rating so we can better understand how we might improve the page.