2 ms·
The robust beauty of improper linear models in decision making
- nkurz 12y agoI can't see any way uniformly weighted models could be better in the general case unless there are strong universal criteria for model creation. But I'd also bet on Gelman's statistical intuition over my own. Let's say I'm trying to come up with a model that predicts the price of a used car based on a list of features: age, color, engine size, brand, condition, etc. Can this possibly be claiming that an unweighted model does better than one that weights "color" less than "age"? I'm particularly interested in this as the IPCC forecasts are done with unweighted combinations of different model results, and this has always struck me as 'broken'. But perhaps it's more defensible than I thought? (looking at the underlying paper now)