I have an ordinal mixed model with four multilevel predictors (it's for exploration). My response variable is ranking 1-4 and so is one of my predictors. My question is if I should treat this ordinal predictor as a continuous variable or a multilevel categorical?

Just some background, in the experiment children selected who they would most like to play with from four photos. The first photo that was selected was given rank 1, the second photo rank 2, third photo rank 3 and the last photo a rank of 4. This is my response variable and the predictor I'm unsure about has exactly the same format but for 'who is most like you?'

I've read loads but almost all I can find is how to treat Likert-scale data and I feel like mine isn't quite that. As far as I can tell I either treat it as a continuous (not do any manipulation) or I recode it into three variables where each ranking is compared to rank 1 ("the best"). Does anyone know for sure?

  • $\begingroup$ I would recode your predictor into 3 categories, as you said, and then include those three dummy variables as predictors in your model (each being compared to whatever you decide is the reference category). Depending on how the variable is coded, you do not have to re-code it. You can tell R it is a factor variable and it will handle it appropriately in the model. $\endgroup$ – Erik Ruzek Jul 23 at 22:22
  • $\begingroup$ I did just that but I used deviation coding the same way as I did for the other multilevel categorical variables. $\endgroup$ – JuliaM Jul 24 at 6:14

Your Answer

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

Browse other questions tagged or ask your own question.