Operations 5: Lies, Damned Lies & Statistics

I have just got back from a week long work based conference, where I went to workshops and symposia to listen to learned professors either teach, or give overviews on their specialist field, and present research findings. I should add at this point that I'm not a researcher, professor, or statistician. My only role in all of this is to understand what the research implies for best practice, and how strong the evidence is.

I mention all of this as a preamble to a few comments that have arisen from various private conversations about the distribution of freight cars across the railroad networks of the North American continent.

When I started in the model railroad hobby the received wisdom was that one should base the number of freight cars per railroad on a rule of thumb that said anything between 25% to 35% would be the home railroad, and then the rest of the freight car models would be based on the connecting railroads, according to their size and proximity to the home railroad.

This paradigm has been replaced by the Gilbert-Nelson model that states that freight cars were seen in proportion to the national fleet. However, I would like to quote directly from Tony Thompson's post a couple of very salient points that people overlook:

"Of course, as they fully recognized, this can only be true of free-running cars like box cars, flat cars and gondolas which are not specially equipped, and is likely true only on main lines. A coal branch, for example, will obviously be quite different.
     They also recognized that certain factors can distort the general pattern. For example, interchange requirements or pool agreements can change the data; so can “hostile” or competitive relations among railroads."

Taking the first paragraph first. My model railroad is based on a traction line that mostly ran passenger services with a few freight trains to service industries along the line. It is therefore not a mainline. I highlighted that part of the above quote to emphasise the importance of the caveat.

The second paragraph's importance is that it defines the confounding variables that can relate to specific presentations of a locale, or railroad.

I have used photographs to support some of the assumptions that I'm making for my freight car roster, because unlike a mainline railroad, freight trains on the North Shore were quite short, and rather than being in the position of being only able to see the first few cars in any train, I often have the luxury of seeing all the cars in the train. I may not be able to fully identify them all by type, but can get close to identifying the owning railroad though.

As for for my aims, I'm trying for verisimilitude rather than quantitative replication. So while I'm aware of the percentages, of freight car types per owning railroad, I'm not going to try and replicate them literally, because I can only have a very small number of models relative to the number required for a statistical representative numbers.

So statistics should be seen as guidelines, rather than rules per se.