You said "It has nothing to do with the number of trains Larry".
My comment about fewer trains raising my percentages was in the context of increasing my sample percentages. Mark Amfahr's data showed that 30 – 33 freight trains per day crossed Sherman Hill in 1949. This implied that the 120 trains in my data represented a sample of about 4.5%, not the "less than 1%" that you had previously asserted. If there were fewer than 30-33 trains per day, then my sample percentages would rise. For example if there were only 20 trains per day my sample percentage would be 7% instead of 4.5%. The sample percentage has everything to do with the number of trains – it is the denominator for calculating the percentage.
And then you said:
"Tim Gilbert pointed out that the percentage of home road cars staying on-line greatly increased throughout the depression years, and so the distribution of cars nationwide was quite different in 1949 than in 1938. The more-or-less uniform distribution of plain box cars is far more apparent in the late 1940's than in the late 1930's. The year 1938 was a severe recession. Industrial output declined sharply. I assume that means there were a lot fewer PRR and other eastern cars on the SP and UP than would be in normal economic times."
Agreed!!! That is exactly what I said in post #89909:
"I think I recall Tim Gilbert writing that the G-N hypothesis fits the data well during economic prosperity but does less well during recessions and depressions. The year 1938 was during the Great Depression, of course. The high proportion of home cars (41%) is another indication that companies may be keeping their cars close by. It might be that modelers wishing to have a realistic mix of cars on their trains should pick an era first – or perhaps even a specific year and season, and then check what was happening in the national economy at that time. The choice of the G-N vs. a regional (or any other) model for an accurate freight car composition may well depend on such ephemera."