Fig. 2: Development of multiple regression models by brute force.
From: Patterns of activity correlate with symptom severity in major depressive disorder patients
A Correlation matrix for all predictors used for model training (training dataset only). The occurrence of individual predictors in the models surviving internal validation is depicted in the right panel. B Performance of models surviving external validation (N = 192) on train and test (independent) datasets. C Cumulative histogram of accuracy (RMSE) achieved by 1 million simulated random datasets with overlayed reference values: (1) average RMSE for models surviving external validation; (2) accuracy level of dummy model; (3) average accuracy of simulated datasets. D Standardized coefficients for all models surviving external validation criteria (N = 192), sorted by RMSE on test dataset in ascending order. Each model is depicted by one column and corresponds to a point in the scatterplot in (B). E Occurrence of most frequently encountered individual predictors the models surviving external validation criteria. “period” and “age” are included based on potential biological relevance.
