Given the research's interest in actual rides carried out by New Yorkers, it was determined that such outlying values should be dropped. The 90% quantile was calculated and trips with duration above this level were dropped from the dataset, reducing its size to 3,120,128.