Why can't I discard natural outliers? [Q]
Say that I have a height dataset given some other variables and there's a guy who's like 8 feet tall. Sure there are people who are 8 feet tall, why should I worsen my prediction of everyone else if the prediction of the guy will be off by a lot as well? It's literally lose lose and if my error metric is quadratic it's gonna skew significantly with outliers.