It looks like everyone else has been adressing how to build and how to control it but not this more important question that you asked.
My problem is that I never decided what to model and tried to model everything that I liked. Not a good plan. Deciding what to model is an excercise in trade-offs. One has to look at all the options and then just let some of the less important ones go because obviously you can't model everything.
So for starters - Do you have a preference on locomotives, steam, diesel, electric? That will help narrow down time and location. Steam obviously means older time wise, unless you want to model a tourist line (most boring) or are satisfied with only 1 or 2 used on excursion trains (also not much fun).
Electric locos were generall used in Chicago area, the northeast, and northwest of the US (obviously there are small exceptions).
Is the interest more in running and operating trains or scenery and watching the trains run through it?
Is there a specific railroad or paint scheme that catches ones fancy?
Is there a certain type of scenery that you want to capture - mountains, hills, plains, swamp, citys, small town, a certain industry that could be the foundation for making a model railroad plan?
Do you want passenger trains that are not Amtrak? That would dictate a time period before 1971. Do you like the old Pullman green w/black roof heavyweights. That would move one further back in time.
I knew a person who built a whole layout around a model of a 1953 Corvette. He wanted it to be new, to be with the top down, in a wide open space, racing the trains, wanted the trains to contrast with the car. He started putting those things together and the plan emerged.
I would recommend making a list with a note on each that indicates how that thing would limit "what to model". With a big list a natural synthesis might occur.
My problem is that everytime I see a cool picture I want to model it. Or when I read another train book I find somehthing in it that I want to model.