The data was cleaned to remove fields other than gender and trip duration, and to remove outliers. The 90% quantile of trip duration was calculated and values above this level were dropped. The cleaned dataset had much more realistic mean and maximum values, allowing the analysis to proceed.