Dissimilarity Index (DI) values per item, over all participants and separately by cohort (* indicates that the solution for DI did not converge so no index value was calculated).

Note: Dissimilarity indices computed for each item represent how well the model under investigation produces expected distributions of response patterns (e.g., from Table 1) that are consistent with observed response patterns for each item. Higher values suggest less consistency between observed and expected values; one recommended cutoff for the index is 0.05 (Dayton, 1998) but this is essentially an arbitrary index value cutoff. Bold items have values below 0.055.*indicates additional convergence problems when DI was computed.†This item was recoded so that all possible points right = 1 and any mistakes = 0.§These items were each assigned one point (i.e., not treated as one point together). Items not represented in this table did not have 0/1 coding (name 3 items), had too much missing data (what floor are we on? What county are we in?) or failed to converge (take this paper, fold it in half) in all 3 groups (and over all responses) so estimates of π* were not computable.