I have collected data in a classification task based on a 3 x 2 factorial design of categorical variables. Participants responded in eight separate tests each, and were asked to group three times in each test - first grouping eight audio stimuli into two groups of four, then further grouping each group of four into two groups of two. The goal is to assess the hierarchical importance of the available factors in decision making. At the experiments conclusion I have a set of eight responses for each participant, which in each case have a primary grouping factor and one or two secondary grouping factors, depending on whether these were consistent or not. I wish to assess how closely these responses fit to a modelled response suggested by meta-analysis of literature, or at least how much of the actual response this model can account for. Regression does not seem to be appropriate in these circumstances since I do not expect to have predictors of response, and I'm not sure whether chi-square goodness of fit is appropriate, nor how to set up the model to perform this test. Can anyone help?