Tim O'Connor said
"You say your data show "dominance" of SP cars -- but isn't it just 201 or so SP cars in 34 random trains over a period of a month? Now, that might show "dominance" but it might just also be random luck. If the UP ran 3 trains a day, I'd say that was a great sample. But even if UP only ran 20 trains a day... Well, it's not much to go on."
The 201 SP cars represent 15.4% of all non-UP boxcars, whereas according to the G-N hypothesis it should be 3.3%. The expected number of cars is 44 according to G-N. That sounds like SP dominance to me, but of course it could be due to "random luck". That would make the many discussions on this list of the presumed anomaly pointless. Perhaps someone more statistically gifted than I am can tell us how likely the dominance is due to the luck of the draw.
The train books cover 120 trains, not 34. (A large number of the trains are Roseville Fruits and other trains with few or no box cars, as Mike Brock has pointed out in the past couple of days.) I wish I could believe they were "random trains" – that would give me much greater confidence that my results are valid and not some sampling fluke.