Prediction of Self Esteem in Internet-based interventions

Models of ordinal outcomes with varying levels of heterogeneity are applied in order to predict the self-esteem of individuals in Internet-based interventions. For this purpose, ecological momentary assessment data is utilized. The parameters are estimated by using Bayesian statistics (STAN) and 10-fold cross validation is applied. The models are evaluated by the RMSE, MAE, DIC, and WAIC.

Setting and other information

• Clone this repository to your computer.
• Change the python directory to the path of the PredictSelfEsteem folder.
• Find the requirements in the requierements.txt file.
• The folder contains multiple scripts for data preparation, application of each model, and visualization of results.
• Raw data data, prepared data (dat_prepared), and example results example_res_hetero_stereo are provided (prepared data is raw data already preprocessed by the prepare_data script and example results are based on the hetero_stereo model).
• Note that the provided data is no real world data. Because of confidential issues, it is not possible to provide real data; the data in this repository is therefore artifical.
• For more accurate results, adjust the executed chains and iteration and warmup period. However, running the algorithms might take a while in order to finish.