Nonprobability purposive, simple random, and systematic sampling was used to collect data. The nonprobability purposive method was used to collect six mukbang YouTubers, three of whom have either self-identified or have been identified through comments as thin, and three of whom have either self-identified or have been identified as not thin. These identities were found using both self-labeling done by the YouTubers themselves and labeling done by commenters that assigned the YouTuber to a weight category. This allowed me to find YouTubers that have enough mukbang videos for a simple random sample, have been uploading mukbang videos since 2016, and have enough popularity to have at least five mukbang videos with over fifty thousand views.
Then, using an incognito browser with the cache cleared and the YouTube search filter “[Name of YouTube channel] mukbang before:2019, after:2016,” a simple random sample using a random number generator was used when taking the top results given by YouTube until 5 videos per YouTuber were identified that have over fifty thousand views. This allowed me to find videos that have enough popularity to be strong for sampling comments while avoiding potential bias towards food types and staying within the popular years for the genre, 2016 through 2019.
From there, a systematic random sample of the top comments on each video was identified, filtering out those that have been posted after 2019 or have under ten likes. These were sampled until fifteen randomly sampled comments from each video were collected, for a total of 450 comments to be coded. By using a systematic sample of the top comments, comments that contain spam or are not popular enough to include as the general idea of the audience were avoided.