0
$\begingroup$

I have a question about how to best analyze data from an experimental psychology study. Briefly, 250 participants were asked to assign four different pictures (Pictures A - D) to two different conditions (Conditions A - B). I would like to know whether Pictures A, B, and C were more likely to be assigned to Condition A relative to Picture D, after adjusting for relevant participant characteristics (e.g., demographic factors). The original plan was to fit a multilevel (random-intercepts) logistic model, in which participants were considered a random factor to account for the fact participants' picture assignments are likely correlated across the four pictures (i.e., within-subject picture assignment correlation). This is the general code I used in Stata (for the crude, unadjusted model).


xtmelogit condition_A pic_A pic_B pic_C || id:, variance OR

However, I have a few questions about the appropriateness of this method of analysis. First, by setting up the model in this way, does this essentially make the sample size within the level 2 groups four (i.e., the four pictures within each level 1, participant, group)? Also, I know that participant characteristics aren't really a random-sample of all possible characteristics, but they are characteristics of the level 2 groups, and thus I included these as level 2 covariates. Is this correct? Thank you so much!!

$\endgroup$

1 Answer 1

0
$\begingroup$

Your code looks right, PotterFan. One thing that is unclear is how the experimental condition enters the model. Perhaps that is condition A or B and hence your interest in contrasting these two choices?

In terms of your level 2 sample size, that is 250 persons. Because each of the 250 persons has 4 data points your level 1 sample size is 250*4 = 1,000. You mention groups in your statement, and I am not sure what you mean by that.

Adding level 2 (person) predictors to the model will reduce the level 2 variance, thus helping you to explain between-person differences in choosing A over B. It is generally desirable to account for person characteristics if you believe those are at all related to choice or you want to explore whether choices vary as a function of person characteristics.

$\endgroup$

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.