Originally Posted by
genejockey
A model has predictive value, or it doesn't. There have been lots of counter-examples, not just one or two. And again, the examples take place in the real world.
You can't say it's representative of the real world, if the real world keeps doing something different.
That isn't true - a model predicts within certain parameters. Outside of those parameters, the model may falter. Yet the model may still be more representative of the real world - particularly since it has greater predictive value than the exceptional anecdotes.