3 ms·

"if you did propose a better model, you'd still have to understand the outdated models in order to prove that they are worse than the model you're proposing."

• copsarebastards 11y ago
  Predictive of what? On what timescales? In which situations? And why? If I'm trying to predict, say, the price of a currency over a day, I'm going to use a different model than if I'm trying to predict the price of a currency over a year, even though it's likely I could use the same models. I'd also pick different models if I knew that oil prices had just gone up 50% than I would if I knew that the S&P had downgraded the US credit rating. Which model you choose comes from an understanding of the inputs of the model, the thinking behind the model, and the situations in which the model is suitable. There's no shortcut here. In any sufficiently complex field, people without any understanding of the field don't even know what they don't know, and therefore can't form valid opinions on the field. Intelligent people whose self-esteem is partially based on their intelligence are particularly susceptible to the Dunning-Kruger effect. Don't fall prey to that. Recognize when you don't know something, and admit it.