Likert Scale Survey Data Analysis in Stata

[Mike Forsch]

Dear all,

I have been reading up on numerous Statalist posts but cannot find the answer I am looking for.

For my thesis, I have conducted a survey with 7-point Likert-scale statements (strongly disagree to strongly agree). This would mean that I need to analyse this as ordinal data since the averages of this would not correspond to one of the textual choices. I found that this requires me to take the median instead of the mean when analysing the data.

My research includes a moderation with two independent variables, two moderators and one dependent variable. The data consists of multiple questions measuring each variable.
Initially, I used the egen and rowmeans functions to generate the new, combined variables. However, I think I need to use the median instead.

Could someone please explain how to best analyse the Likert-scale data I gathered?


Thank you in advance!


Mike

[ericmelse]

Dear Mike,

Your question is 'big' in the sense that there is much literature on this subject.

My best recommendation is to buy this Stata book A Course in Item Response Theory and Modeling with Stata from the bookstore.

It is a fine introduction into IRT-modelling as an alternative for (just) summing up item scores (either as a mean or a median, something I rather would only do as a secondary analysis).

[Rich Goldstein]

Likert scales sometimes have issues re: the middle point - its meaning can be ambiguous; you might want to take a loot at Tutz, F (2021), " Hierarchical models for the analysis of Likert scales in regression and item response analysis," International Statistical Review, 89(1): 18-35