Instructions
Create a CSV file with three columns:
- y - the actual value of the variable you want to predict. For 'AUC' and 'Balanced accuracy' estimates, it must contain binary values. For 'cor' and 'mse' estimates, it must have numeric values. For estimating the hazard ratio of the predictions in survival data, instead of creating the 'y' column, create a 'time' column with the times and a 'status' column with status (1 for events, 0 for censors).
- y.pred - the predicted value of the variable. For 'AUC', 'Correlation', 'Hazard Ratio', and 'MSE' estimates, it must be a numeric vector. For 'Balanced accuracy' estimates, it must contain binary values.
- site - the name of the site for each observation.
Afterward, select the estimate and the CSV file in the left sidebar.
Citation
Solanes, A., Palau, P., Fortea, L., Salvador, R., Gonzalez-Navarro, L., Llach, C.D., Valenti, M., Vieta, E., Radua, J. (2021) Biased accuracy in multisite machine-learning studies due to incomplete removal of the effects of the site. Psychiatry Research: Neuroimaging, 314:111313.