I want to do an analysis on word search performance.

Word search performance is my dependent variable. A word puzzle was assigned and the correct number of words is the score, for example 5 correct answers means the person got a score of 5. However for easier control, i had ranged the score to categories. That is if a person got 1-4 then low performance, 5-8 medium performance and 9-13 high performance.

My question is since now i have categorised the score into 3 levels. Do i have to apply an ordinal logistic regression or may i use a multinomial logistic since the order of word search performance doesn't really matter?

  • $\begingroup$ What does "for easier control" mean? If you have the original data for the performance, i. e. the metric number of correct answers, in 9 out of 10 cases (or more?) you should really model that number, even if later downstream only the performance category is relevant. $\endgroup$ – Bernhard May 29 at 21:39

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy