# Table of Contents #
-------------------
This repository contains supplementary information related to the article "Personality Computing with Naturalistic Music Listening Behavior: Comparing Audio and Lyrics Preferences".
> **In-Text Figures:**
All Figures contained in the original article under a CC_BY4.0 license
> **Online Supplemental Materials:**
**Supplemental Methods:**
- **Disclosure_of_Prior_Data_Uses.pdf**: Disclosure of all prior publications and a preliminary preregistration based on (parts of) the same data.
- **Lyrics_Preprocessing.pdf**: Detailed descriptions for the
preprocessing of the textual song lyrics data
- **TABLEX1_predictor_definitions.pdf**: List and definition of all behavioral predictor variables
**Supplemental Results:**
*Descriptive Statistics:*
- **TABLEX2_predictors_summary.csv**: Summary statistics for all music listening (i.e., predictor) variables
- **TABLEX3_predictors_inter-correlation.csv**: Pairwise Spearman correlations between all pairs of music listening variables (i.e., predictors)
- **TABLEX4_personality&demographics_corr.pdf**: Pairwise Spearman correlations (incl., the 95% CI) between Big Five scores (i.e., outcomes) and demographic variables
- **TABLEX5_personality&predictors_correlations.csv**: Pairwise Spearman correlations (incl., the 95% CI) between Big Five scores (i.e., outcomes) and music listening variables (i.e., predictors)
- **TABLEX6_personality&predictorstop_TOPcorrelations.pdf**: Highest pairwise Spearman correlations between Big Five scores (i.e., outcomes) and music listening variables (i.e., predictors)
*ML Modeling:*
- **FIGUREX1_performance_MSE.pdf**: Box- & Whisker-plots of
the full benchmarks based on the MSE evaluation metric
- **FIGUREX2_grouped_performance_rho.pdf**: Box- & Whisker-plot of the random forest models' grouped benchmarks based on the Spearman correlation evaluation metric
- **TABLEX7_grouped_performance.pdf**: Aggregated performances of the benchmarks comparing the different variable groups for random forest and elastic net models and different evaluation metrics
- **TABLEX8_variable_importance_RF.csv**: Variable importances for all predictors in random forest models with a prediction accuracy of rho >= .10
- **TABLEX9_betas_EN.csv**: Beta weights for all predictors in elastic net models with a prediction accuracy of rho >= .10
> **Open Code & Data:**
Download and unzip the **"PersonalityComputingwithMusic.zip"** file to re-run our analyses. Set the working directory to this folder and run the different analysis steps in the correct order indicated by the numbering.
We provide the entire code for the different steps of analysis from obtaining external data to variable extraction to prediction modeling.
However, we can only provide the aggregated data after the variable extraction step. Raw logging data and raw song-level data (i.e., Spotify features & song lyrics) cannot be provided openly due to privacy and copyright issues. Hence, only the code for the analyses (starting with "6.descriptives") will run on your computer but not the preprocessing code.
Finally, the folder contains the R objects for the different steps of analysis that were created by the respective code.
A detailed explanation of the folder tree after unzipping can be found in the different READMEs.
