Table 2.
FTP Linear Mixed-Effect Models
| Model 1 | Thickness ~ IT FTP ✕ Time + Covariates + (Time|Participant) |
| Model 2A | Thickness ~ IT FTP ✕ Time ✕ G - G - IT FTP:G + Covariates + (Time|Participant) |
| Model 2B | Thickness ~ IT FTP:Time:G1 + IT FTP:Time:G2 + Time:G1 + Time:G2 + IT FTP + Covariates + (Time|Participant) |
Model 1) IT FTP by time interaction across all MR timepoints; Model 2A) To see slope differences in retrospective and prospective analyses, the two time periods are forced to have a shared intercept at time 0, which is achieved by including a three-way interaction with observation period (G) which codes whether an MR time point was obtained before or after FTP-PET, and then removing intercept terms G and IT FTP:G; Model 2B) Statistically equivalent to Model 2A, this was used to extract retrospective and prospective estimates directly. G1 is an indicator variable for the retrospective observation period and G2 is an indicator for the prospective observation period. Note: Wilkinson notation is used, ‘✕‘ indicates an interaction with all combinations of lower-order terms, while ‘:’ represents an interaction without lower-order terms (e.g., x1 ✕ x2 ✕ x3 = x1:x2:x3 + x1:x2 + x1:x3 + x2:x3 + x1 + x2+ x3). As noted in the text, both models included ROI thickness as dependent variables, age and sex by time covariates, as well as random participant intercepts and random time slopes. Abbreviations: FTP = flortaucipir; IT = inferior temporal gyrus.