Kevinstat
Jr. Member
Posts: 1,823
|
|
« on: September 15, 2020, 06:26:01 PM » |
|
|
« edited: September 28, 2020, 07:26:04 PM by Kevinstat »
|
This is something I first thought of years ago, but never actually asked for until a few weeks ago, when I sent Dave Leip an e-mail through the form on the Atlas with this request. I never heard back, but when I saw that Dave's been active on the forum (even though it's now a separate site, which I assume has to do with the whole copyright, fair use legal threats thing), I figured I'd put in the request here.
On any of the prediction pages, you can click on a state to see how many people predicted which candidate to win, with which "percentage decade," and the level of confidence they felt in their prediction. The median predictions are also shown on the maps for the states and (for the general election Presidential predictions going back to 2004, for Maine and Nebraska's congressional districts). For the states (but not the CDs) you can hover over the state to see what the median is (in case you're not sure you're eyeballing the color shades correctly).
But aside from the median information from the color/shade on the map (and there's not even the hover-over info there, although it would be difficult to do that due to space constraints and it wouldn't really be needed if my main request was granted), there's no such information for congressional districts, either on their own pages or as additional information on the state pages. I was hoping that could perhaps be added for 2020 (I don't see any point in going back to 2004 for that). It seems like a logical extension of both allowing people to make predictions for those CDs (I'm sure you'd take a lot of grief if you didn't do that, given the EV allocation) and your letting people see the predictions at the state level.
I'm also curious about how ties are treated in determining the median (like if there are two different "co-medians"). Do you round ties between two candidates in favor of the candidate who last won the state/CD? Do you use diagonal stripes? And what about different percentage decades or confidence intervals for the same candidate (or leaning for a candidate vs. being a tossup)? How do you round that. Ties are probably fleeting but I would think they have to happen at least once whenever there's a change in the median, and looking at the Aggregate Prediction History for 2020 there have been a whole bunch of changes this cycle.
Thank you for your consideration, and thanks in advance for whatever you are able to add and whatever information you are able to give on my question re: ties.
|