I'm doing research for a class, and I'm collecting the following data:

enter image description here

The background questions are my independent variables and the survey questions are my dependent variables. For the latter, I'm using a 5-point Likert scale (strongly disagree, disagree, etc.)

The purpose is to observe if there's a difference in preference for food in different demographic groups. E.g. People with higher education have a high preference for Italian food, older people prefer Chinese food than younger people, etc.

Some of my hypotheses:

  1. There is a difference in preference across age, income, education, etc.
  2. People with higher education have a higher preference for Japanese food than those less educated.
  3. People with middle income have a higher preference for Italian food than for other cuisines.

I'm struggling a bit, because the only hypothesis testing we've learned is t-test and ANOVA, and I don't think I can use those on ordinal data. I've googled a bit, and I've come across two potential tests I can use:

  • Ordinal regression
  • Chi-squared test

Do you think these can be used to test my hypotheses? Hope you can point me to the right direction.


1 Answer 1


When you have only ordinal Y then ordinal regression can be ideal. If you have only a single X that is possibly ordinal then rank regression may be what you need, with careful attention to how ties are handled. Some choices include Somers' $D_{xy}$, Goodman-Kruskal $\gamma$, and various versions of Kendall's $\tau$. When you have more than one X and want to respect the ordinal nature of one or more of the Xs then you may need to use a Bayesian model that is tailored to ordinal X such as what the R brms package does by default for ordered factors.


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.