# Integrating ecological momentary and passive sensing data to improve depression severity prediction: insights from the WARN-D study
Project hub for the associated manuscript containing relevant necessary information for replication.
--------
- **Manuscript:** Includes the manuscript file, figures, and the supplementary materials text.
- **Data:** Contains participant ID's, final models generated by the pipeline, and files containing the final results of used for statistical analysis and figure generation. Naming conventions indicate whether models were predicting 1) daily or weekly depression, 2) used the EMA, passive sensing (EPA), all data, or the lagged PHQ data and 3) which fold they were trained and validated on.
- *keras_models.zip*: Final models generated by the training procedure across all folds. Model names consist of the base `model_{day/week}_{ema/epa/all/phq}_fold{1-10}.keras`
- *training_test.zip*: CSV files with the final results of the 10-fold CV process per fold, and the predictions yielded on the test set for each of the validated 10-fold models. Base naming `results_{training/testing}_{day/week}_{ema/epa/all/PHQ}_fold{1-10}.csv`
- *shap_values.zip:* SHAP values estimated in python, and saved for plotting in R because GGPLOT makes prettier plots. Base naming `shap_values_{day/week}_{ema/epa/all}.csv`
<br><br/>
- **Scripts:** Scripts used in various parts of the study. Contains two subdirectories. [ALICE](https://pubappslu.atlassian.net/wiki/spaces/HPCWIKI/pages/37519378/About+ALICE) contains scripts run on the Leiden University computational cluster, while **LOCAL** contains scripts run on local device primarily for data generation, applying models to test data, generating SHAP estimates, and figures.
- ALICE:
1. `bash_hb_{day/week}_{ema/epa/all/phq}.sh`: Wrapper scripts to submit the hyperband tuning scripts for each of the corresponding models to the SLURM cluster (see 2).
2. `model_hb_{day/week}_{ema/epa/all/phq}.py`: Scripts to run hyperband tuning algorithm using KerasTuner on the computational cluster.
3. `bash_final_{day/week}_{ema/epa/all/phq}.sh`: Wrapper scripts to run the python scripts for the final models based on the hyperband tuning results (see 4).
4. `model_final_{day/week}_{ema/epa/all/phq}.py`: Scripts to run validation models with 10-fold CV using the final tuning parameters derived from step 2.
- LOCAL:
1. `notebook_dataprep.R`: Used to explore the data, generate descriptives, and save the data in a format needed for the LSTMs in python. Output from this notebook is used on the ALICE cluster (see above). Run prior to ALICE step 1.
2. `notebook_results.ipynb`: Jupyter notebook run locally on the models saved from step 4 in ALICE on the test data.
3. `notebook_figures.R`: R Notebook for getting the summary statistics, testing of models, and making of figures.
