--- In STMFC@..., "Mike Brock" <brockm@...> wrote:

Actually I said:

"The model that I prefer is a modified Nelson/Gilbert model which

states

that
RRs with "significant interchange" should have from 2 to 2.5 times the
national %."

I'm sure this is the key… I'm sure that after one applies a whole
bunch of "weighting factors" that account for proximity of other
roads, preferred interchange partners, preferred routes for "rollers",
etc, the little bit of traffic that's left will look quite similar to
the Gilbert / Nelson proportions. The problem is, those weighting
factors are going to be different for every stretch of railroad one
could possibly model. As examples, compare the Yosemite Valley, which
had almost no cars of its own, with the similar sized Greater Winnipeg
Water Works District Railway, which had, as far as I know, no reason
to handle a foreign car, since the line basically functioned as a
conduit to bring gravel from pits along the line into the city for use
by the local construction industry.

I would suspect that the closest to the "average" stretch of railroad
would be the NKP or Wabash; railroads smack dab in the middle of the
country that handled the largest proportion of overhead traffic vs.
loads originating and terminating on line. If one had good train
consists for those lines, perhaps that would be the place to start
trying to determine correction factors for proximity and connections
of the "average" railroad.

But why bother? Data for the average railroad is only going to be good
for someone who freelances, but even then you run into the problem
that if you model the Maumee, what the heck is the Wabash hauling if
all that traffic is on the Maumee? Even if you get what was actually
happening right, you then have to modify it again to take into account
the new player you added.

I personally think that time would be better spent studying the
prototype one is trying to model, identifying the consists of the
trains as best one can from consists, interchange statistics, photos,
movies, whatever is available, noting not only the overall car mix,
but specific instances of heavy concentrations, because those heavy
concentrations aren't random events, they MEAN something, and modeling
them helps to capture the feel of the prototype.

Dennis