This repository contains the preprocessing pipeline from the BOKI study. The scripts have been developed by me, Irene Sophia Plank, at the NEVIA lab (https://nevialab.com/). Some of the scripts are inspired by or further developed from https://github.com/IreneSophia/ML_speech or Jana Koehler's https://github.com/jckoe/MLASS-study. The study has been preregistrated on OSF (https://osf.io/7jtzr). It extends efforts of finding diagnostic markers for psychiatric disorders with symptoms affecting social interactions to Borderline Personality Disorder.
The data was collected at the Medical Faculty of the LMU Munich. Strangers were paired in either heterogeneous (one interaction partner with and one without BPD) or homogeneous (two interaction partners without BPD) dyads and asked to have two ten-minute conversations, one about their hobbies and one to collaboratively plan a menu with foods and drinks that they both dislike. We collected audio, video and psychophysiological data during these conversation. Previously, we have collected similar data with heterogeneous dyads including one autistic interaction partner (Koehler et al., 2022: https://doi.org/10.1101/2022.12.20.22283571; https://github.com/IreneSophia/ML_speech).
The folders contain preprocessing scripts for the different modalities as well as synchrony computation within and across channels. For the synchrony computation, the rMEA package and specifically MEA objects and the CCF function which computes windowed lagged cross correlation are used. Some of these scripts rely on already preprocessed data, e.g., with OpenFace or MEA. Details can be found in the individual folders.