You are correct as far as you go, but what you're saying is contained in the first ten pages of any book on device reliability. There is at least one whole Ph.D. project's worth of additional research between your statement and the time when you throw up your hands in despair because your Monte Carlo models just don't seem to work.

If you really want to know, you'll have to read the papers. I'm sure there are quite a lot of them. The incentive to solve this problem is very, very high.