RegionEx defense against MDA
Dear Ms. Lente,
Upon doing my own analysis I have discovered that your comments about average flight delay and percentage of delayed flights by RegionEx as compared to MDA are not quite correct, or at least do not tell the whole story. First, RegionEx’s % of delayed flights is 26.3% compared to 26.5% by MDA, a small outperformance but better nonetheless. Second, RegionEx’s average delays are significantly skewed by some large 100+ minute delays (refer to RegionEx’s scatter plot), making the 5 minute deficit in the mean a deceptive data. This deception is also reflected by the very large standard deviation observed in RegionEx’s descriptive statistics. A quick scan of the medians and modes actually shows that more often RegionEx is performing a bit better than MDA. A quick scan of the histograms indicate that while 57.5% of RegionEx flights are less than 10 minutes delayed, MDA flights are more than 10 minutes delayed 72% of the time, despite the 10 minute additional cushion it schedules for every flight as compared to RegionEx scheduled flight times (100 minutes for MDA, 90 minutes for RegionEx). This means that if comparing apples to apples and giving a 100 min allowance to RegionEx flights, 57.5% of RegionEx flights would register as being on time, a vast improvement over MDA. While the fact that 24% of RegionEx flights are more than 15 minutes delayed is indeed concerning, this data as an aggregate seems to be a representation of Simpson’s Paradox, where the aggregate result looks worse than the data broken down individually.
Thank you,