I am running an ordinal regression for an experiment I have conducted, where the DV is ratings on a 5-point scale, and the predictors are the type of item, the type of participant (native speaker/L2/heritage) and the task (there were two tasks, both targeting the same type of linguistic phenomenon). I want to add random effects for participant and for item. I added an intercept for item (1|item).

The problem: How to add the random effect for participant? There were different lists in the experiment, meaning - every participant only saw a subset of all items, no participant saw all items. I feel that this means that I cannot just add an intercept for participant, but have to somehow use different syntax to reflect the fact that each participant only saw a subset of the items.

There isn't a column/variable for the list number, so I cannot use participant as a variable nested in that.

How should I add a participant random effect?

  • $\begingroup$ Could you make each combination of items that the individuals were exposed to as a factor, e.g. "A+B+C", "A+C", and then use that factor as a random effect? $\endgroup$
    – mhovd
    Dec 25 '20 at 15:03
  • $\begingroup$ I can do that. How should I phrase that random effect, then? Would it be (1+item|combination)? $\endgroup$
    – maria
    Dec 26 '20 at 20:33
  • $\begingroup$ or maybe it should be (1|participant:combination)? $\endgroup$
    – maria
    Dec 27 '20 at 0:18
  • $\begingroup$ I believe they should be separate, e.g. y ~ (1|combination) + (1|id) + item. However, it depends on your experimental setup and hypothesis. $\endgroup$
    – mhovd
    Dec 28 '20 at 11:02
  • $\begingroup$ thanks, I did decide on doing that, what you proposed above $\endgroup$
    – maria
    Dec 28 '20 at 21:27

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.