Additional file 7 of Video-based long-term follow up of musician’s dystonia in pianists reveals similar improvements following different treatment strategies: a retrospective observational study

Additional file 7. Leave-one-out cross-validation estimates (via expected log pointwise predictive density, elpd) of regression models estimating dystonia-related movement impairment while playing c-major scales on the piano keyboard. The model with the largest elpd is deemed to be the best model and listed at the top, and the other models are listed by decreasing epld, with the column elpd_diff showing their ‘distance’ to the top model. The column se_diff contains the pairwise standard error of each model with the top model. If a model lies more than approximately two SE away from the top model it is considered to be meaningfully worse.