Methodology

Some counties in the US, like Los Angeles or New York, have really high populations. Others are sparsely populated, like North Slope Borough in Alaska, which has one person for every 10 square miles of land. In order to control for population effects — to make sure our maps fairly indicated regional interest, rather than places where people watched the most YouTube videos — we looked at each movie’s trailers, and calculated what percentage they constituted of all US YouTube views in the film’s opening week. This gave us a national baseline. We then ran the same analysis for each county. Finally, to get a measure of county interest, we looked at the degree to which each county’s score exceeded or fell below the national average. Prior to mapping, we smoothed the data by calculating Getis-Ord Gi* z-scores for each county’s 20 nearest neighbours. So, to recap in English: we looked at whether a county’s viewing habits for each trailer were higher or lower than the rest of America.