Its not just not a good proxy for all potential private models. Its not even a good proxy for a model set up pretty much like the American system, or even an attempt to exactly duplicate the American system as close as possible. When you dealing with complex systems (health care and health care insurance systems) that are in turn greatly effected by many variables outside the system itself (many economic factors, diet and exercise, genetics, environmental factors, social, cultural and psychological factors, accident rates, etc. etc. etc.), a sample of one is no good unless you can nearly exactly duplicate, or perfectly control for the other factors (both of which are probably impossible)
This might get a response like "we can't have a sample of thousands, we have to deal with what we have", but not being able to get good data, doesn't change the poor data in to good data. A sample of 12 on one side is pathetic, a sample of one on the other is useless. |
