3
$\begingroup$

I am a linguistics graduate student, and I'm a bit lost regarding statistics. I'm performing a quasiexperiment, where I have three groups separated using a categorical variable (voice heard). Every person in each group listens to 40 sentences, which can be categorized according to two variables, predictability and intelligibility, also categorical variables (both can be just low or high). With respect to these last two variables, there is an equal number of sentences for each combination. After the test is taken, I'm computing the total score (from 1 to 10) for each combination of sentences. The number of people in each group is different.

What I'm interested in is determining which one of the variables (or combination of them) is significant to the scores. I've read about ANOVA (1-way and k-way), Kruskal-Wallis, ordinal logistic regression, but it seems that none of them is really appropriate for my setup, a discrete variable depending on categorical variables.

Improve this question
$\endgroup$
2
  • $\begingroup$ Welcome to the site. Could you provide a snapshot of what your data table looks like? Many of us are data-oriented and many are not native English speakers, so providing even some toy data if you don't want to share your real data would greatly help those who can help you with your problem. $\endgroup$
    – Marquis de Carabas
    Commented Sep 27, 2017 at 14:30
  • $\begingroup$ I'm also having trouble distinguishing between discrete and categorical variables, since they are interrelated. Do you mean a count and categorical? $\endgroup$
    – Marquis de Carabas
    Commented Sep 27, 2017 at 14:36

1 Answer 1

Reset to default
3
$\begingroup$

Why do you think that ordinal logistic regression isn't appropriate for your setup? It seems to fit your description quite well. Often, it doesn't really matter whether your predictors are categorical or continuous variables, especially if we are dealing with binary variables as it's the case here. If your variables are non-binary, it can get a bit tricky to code them appropriately. Anyway, there shouldn't be a problem to run an olr on categorical variables. Your 1to10-Score dependent variable should fit olr quite well to. As you noted, it is not a continuous but discrete, ordinal variable.

Improve this answer
$\endgroup$

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.