I am examining a class that did 3 maths questions which required explaining their answers, at two different time points. Their answers are going to be coded by,

1) answered correctly

2) used the correct method but wrong answer

3) used an incorrect method

4) did not answer

So I have a categorical variable at two time points. I want to see if the participant's category at time one per question predicts the category at time two for the same question.

In order to do this I'm wondering if I can use multinomial logistic regressions, where time one is the independent variable (categorical, 4 categories) and time two is the dependant variable (categorical, 4 categories). And run this individual for the three maths questions.

I'm wondering if I am on the right track, if this statistical analysis fits my question and if it is within the assumptions of the analysis?

Many thanks