Statistical analysis of rapid prosody description

• Study design: Stefan Baumann
• Data collection & preparation: Stefan Baumann & Janina Kalbertodt
• Statistical Analysis: Bodo Winter

Libraries required for this analysis:

• lme4
• party
• DMwR
• dplyr
• reshape2
• xlsx

Script files contained in this analysis:

• 001_RPT_individual_analysis_preprocessing.R
  The main preprocessing script, works on Excel files and outputs tidy csv files.
• 002_mixed_model_analyses.R
  Computes mixed models (but does not interpret and visualize them). Warning: Takes a lot of time to run.
• 003_random_forest_analyses.R
  Computes random forests and variable importances (but does not interpret and visualize them). Warning: Takes a lot of tim to run.
• 004_random_forest_visualization.R
  Interprets and visualizes random forests.
• 005_visualizations.R
  Interprets and visualizes mixed models and other analyses.

Data files contained in this analysis:

• rpt-Daten-15juli2015.xls Contains all summary data, that is, prominence score averages (overa all listeners) for each word
• rpt-Daten-31juli2015_spectral_tilt.xls This is the most up-to-date file of the summary data
• rpt_Einzelwerte-25juli2014-1.xls Contains individual level data, that is, all prominence ratings from each listener (wide format)
• RPT_summary_processed.csv The summary data, cleaned and in English.
• RPT_individual_processed.csv The individual level data, cleaned (long format) and in English.
• CODEBOOK.md description of all columns.
• listener_gender_info.csv is needed to map genders onto listeners.
• speaker_gender_info.csv is needed to map genders onto speakers.
• block_order_information.csv is needed to map block orders (there were two block orders) for each participant.