Hacker News Reader

This was a nice article, but can anybody reference a longer, more mathematical analysis?

At least here they talk about weighting the polling results, but I would really like to know how, exactly.

This is not a decent answer to your question, but maybe something can be learned from the analysis company's website.

I guess they're pretty unlikely to give us much detail about their analysis models, secret sauce, proprietary approach, etc, but maybe there's something mildly interesting here:

https://cambridgeanalytica.org/datamodels

Cambridge Analytica claims to have a database of over 220 million Americans.

There are two parts to a polling prediction: predicting how each demographic is going to vote, and predicting how many of each group will show up to vote. I believe the weighting refers to the second part of the process. It appears that the groups more likely to vote for Hillary stayed home on Tuesday more than anybody predicted.

There's an article I read within the last couple of days that goes into detail about the weighting process, but I couldn't find it again.

Reply on news.ycombinator.com