Multilevel binary logistic regression assumptions

Question

I am struggling to work out the assumptions of hierarchical binary logistic regression, to test whether my data is suitable for such an analysis.

My data is repeated measures (each participant ID consists of 64 rows of data (1 per trial) where the IVs are manipulated each trial. The IVs consist of 1) cue location and 2) cue congruency, where each has two levels/binary. Each participant experiences (in a random order) 16 of each combination of the IVs (cuetop-congruent, cuetop-incongruent, cuebottom-congruent, cuebottom-incongruent). The DV for all trials is stimulus-location (binary: either top or bottom).

I am testing whether 1)cue location predicts stimulus location and, more importantly, 2)cue location*cue congruency interaction is significant/can predict stimulus location.

First, is a hierarchical(multilevel/mixed model) binary logistic regression the right anaylsis to use? If so, what are the assumptions I have to check to ensure the data fits the statistical test requirements? For example, I am unsure whether a normal distribution is needed, and if so, which variable/s does this apply to? I am clueless about the assumptions so as much detail as possible would be appreciated!!

Answer 1

However, in a multilevel logistic model, many different residual plots can be used to inspect model assumptions. In the multilevel logistics model, the level-two variance should fulfill the assumptions of linearity and normality. Since the distribution of the responses was binomial, only the level two residual properties were examined (Snijders and Bosker 2011). Snijders, T. A. and R. J. Bosker (2011). "Multilevel analysis: An introduction to basic and advanced multilevel modeling." Sage.