3 ms·
There is a proliferation of predictive models and their outcomes vary widely. LA Times consistently predicted a Trump win, FiveThirtyEight was down to a coin to
by andr 10y ago
There is a proliferation of predictive models and their outcomes vary widely. LA Times consistently predicted a Trump win, FiveThirtyEight was down to a coin toss a few times, while The New York Times was persistently pro-Hillary. While those models were meant to help people make sense of the large number of individual polls, now we need models to make sense of the models.
- dj-wonk 10y agoYes, and this is a good thing. It is better than what we do, informally. We all use models (of some sort) when we process information; e.g. I discount certain publications and value others. But few people actually make their models explicit and ask others to critique them. Making a model public and explicit is a great first step. Building more models on top is a good thing, not a bad thing.