![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
ciphergothThat last chart looked pretty favourable to Clinton. Was that an accurate impression? The last chart showed Clinton winning Florida by eight points and Obama losing it by one point. The trouble is that these polls have an error margin of four or five percent just through sampling variation, and that's leaving aside the errors due to the unpredictability of turnout and straightforward drift of the electorate due to people changing their minds. How can we take that into account?
What I've done here is run lots of simulated elections. In each election, we assume a certain error rate in the polls; we disturb all the polls by a random amount based on the error rate, and work out who wins. Each data point is the result of 10000 simulated elections.
The conclusion is: If you think that the overall error in the polls is likely to come down to 6 points or less, vote Clinton; if you think it's going to be more than six points, vote Obama.
Update: after reading this post, why not fill in the poll?
What I've done here is run lots of simulated elections. In each election, we assume a certain error rate in the polls; we disturb all the polls by a random amount based on the error rate, and work out who wins. Each data point is the result of 10000 simulated elections.
The conclusion is: If you think that the overall error in the polls is likely to come down to 6 points or less, vote Clinton; if you think it's going to be more than six points, vote Obama.
Update: after reading this post, why not fill in the poll?
