I ran an experimental study with a within-subjects design. Subjects learned words under 4 different conditions (A,B,C,D), 30 words in each condition. Shortly after, they did a yes/no recognition test for these 4 item types, mixed with new unseen words. I also measured another variable (X) that I thought might predict recognition, which is continuous and is a single value for each participant. I am specifically interested in recognition (y) of items from conditions A and B, and I found that X significantly predicts y(A) but not y(B). I see two problems with it:

1. I have a repeated measures design, so y(A) and y(B) will have correlations since they are from the same subject.

2. I did not specify the model to include all y(A), y(B), y(C), and y(D), which means that

Normally I'd run a repeated measures ANOVA with item type as the IV and recognition as the DV but with a continuous outcome it seems to make things more complicated. I think I need to use some sort of a multi-level model for this but I have no experience with it, nor do I know how to implement it. I am struggling to find information about how to move forward with this, so I could use some feedback/discussion.