The repository included an extended* version of the Stan model (which also had an indexing bug).
While testing extended versions of the model ultimately used for the paper, GB had unfortunately saved that extended model under the same filename as the model ultimately used for the analysis. This remained undetected, because compiling the R-Markdown document does not include fitting of the time course model, because this takes several hours.
*The extended model also used the variable ld_centre, which codes in which test centre category a subject had the longest stay, to model variations in the intercept (peak load) and the down-slope. This added complexity did not lead to a better model-fit and was therefore not included in the final model.
The repository included an extended* version of the Stan model (which also had an indexing bug). While testing extended versions of the model ultimately used for the paper, GB had unfortunately saved that extended model under the same filename as the model ultimately used for the analysis. This remained undetected, because compiling the R-Markdown document does not include fitting of the time course model, because this takes several hours.
*The extended model also used the variable ld_centre, which codes in which test centre category a subject had the longest stay, to model variations in the intercept (peak load) and the down-slope. This added complexity did not lead to a better model-fit and was therefore not included in the final model.