GIGO
Posted: Tue Oct 15, 2013 11:31 pm
It can't be said often enough, do not blindly trust a computer model:
http://drtimball.com/2011/computer-mode ... n-society/
It's just too tempting and all too easy to play games with coefficients and abstractions to mimic behavior and then trust that the variables will remain the same. The problem with that surprise free scenario is that when things go out of bounds you can't say if it's significant or not. I've seen too many models that do exactly that. Actually made one myself. Fortunately it was just a college project where nothing was at risk other than a grade. But I did learn that you have to watch computer models very closely.
http://drtimball.com/2011/computer-mode ... n-society/
It's just too tempting and all too easy to play games with coefficients and abstractions to mimic behavior and then trust that the variables will remain the same. The problem with that surprise free scenario is that when things go out of bounds you can't say if it's significant or not. I've seen too many models that do exactly that. Actually made one myself. Fortunately it was just a college project where nothing was at risk other than a grade. But I did learn that you have to watch computer models very closely.