It is a fascinating project. I have pondered other factors that would be interesting to investigate. For example, Boston is unique in that it is the only major that has an industry built around getting runners their BQ. That has created a number of marathons (mostly downhill) that actively promote their BQ% as an enticement. I run a local downhill to get my Boston entries and I have calculated that it shaves off about 14 minutes from my finish time. That is, it is far easier for me to BQ on a carefully selected feeder race than it is to BQ in Boston. It may be possible - with a lot of web scraping - to include this information in your dataset. Also, I only began running marathons when I retired. There is no way I would have had the time to adequately train when I was working. The older demographic, of course, skews toward retirement which may represent a de facto "softer" entry. Similarly, the older folks might have the time and the means to travel to a high BQ% qualifier than younger folks would.