1. Raw video data to dataset processing: 4 example video files so that feature_final.py can be run to produce CCR_final_nc_visual.csv (for our paper, we ran feature_final.py on all video files to generate the video-related features of CCR_final_nc_new.csv; all other columns of CCR_final_nc_new.csv were directly obtained in the data gathering process or based on human coding).

2. Replication files: replication data, CCR_final_nc_new.csv, and code, CCR_RR_nc.R, to produce all figures and tables in The Pervasive Presence of Chinese Government Content on Douyin Trending Videos (Computational Communication Research)

Variables from metadata
create_date: creation date of the video
create_time: creation time of the video
duration: video length
topic_name: trending topic that the video relates to
account_hash: hashed id of the account

Variables extraced from videos
frame_numbers: total number of frames contained in a video
frame_numbers_sampled: total number of frames used for analysis
luminance_avg: video brightness
entropy_avg: video color complexity
face_binary: number of frames containing faces
warmth: warm color dominance score
cold: cold color dominance score
face_rate: proportion of frames that contain faces

Variables from human coding
account_type: type of the account
regime_acct_type: further disaggregates regime-affiliated accounts
topic_category: category of the trending topic
covid: whether a topic relates to covid-19

@article{lu2022pervasive,
  title={The Pervasive Presence of Chinese Government Content on Douyin Trending Videos},
  author={Lu, Yingdan and Pan, Jennifer},
  journal={Computational Communication Research},
  volume={4},
  number={1},
  year={2022},
  pages={68--97},
  publisher={Amsterdam University Press}
}