2016 U.S. Presidential Race Megathread Mark 2.

Zubeneschamali wrote: »

It is the state-by-state electoral college numbers that tell which prediction is better.

My earlier post(s) were admittedly relaxed, descriptive, and not inferential consistent with the spirit and intent of this discussion forum. If someone wanted to know state-by-state if forecasts by comparative organisations were better or worse, a simple intuitive eyeballing, tabling, graphing, or elementary arithmetic computation perhaps with percentages may fail to suggest with any rigour if they were better or worse, given the limitations and problematic nature of such intuitive or elementary approaches; e.g., random variability, different types of error, confidence levels, etc., would not be accounted for and could lead to completely spurious conclusions (see Darrell Huff in "How to Lie with Statistics" for a fun read).

Someone may elect to use a bit higher level of analysis to see if there might be statistical differences between comparative state-by-state forecasts, perhaps using a nominal win/loss by organisation forecast chi square matrix at the p<.05 level of significance? Or perhaps raising the analysis a step higher to the ordinal level by taking into account the light to dark blue, red, and grey colours of the states appearing in the 50 state forecast maps by organisation and subjecting it to an ordinal test of significant differences? Of course, both nominal and ordinal data treatments would be nonparametric and subject to the appropriate cautions when interpreting the results. At first blush it's doubtful that we could raise this comparative analysis to the parametric level acknowledging the varying degrees of subjective opinion evident between different organisations and their state-by-state colour assignments (but after a bit of needed sleep, I might think of creative ways to accomplish this).

In concluding, yes, I know that my earlier comments were relaxed and descriptive and not inferential, but who wants to go into such analytic detail when commenting on this election? Then again, if you are having trouble getting enough sleep, my above statistical rambling may be helpful? And if you are really in need of sleeping help, you are encouraged to visit and post in our Sleeping & Dreaming forum. This makes me YAWN just thinking about it.

**Apologies Zubeneschamali. I'm wired by tired after a work all-nigher, and way too much java!** :eek:

Zubeneschamali wrote: »

For the really OCD out there, live updates on battleground states:

http://www.slate.com/votecastr_election_day_turnout_tracker.html

Right now, Clinton winning Colorado, Florida, Iowa, Nevada, Ohio and Wisconsin, Trump winning Pennsylvania.

Bloody Sylvanians.

It's way, way too early in the state-by-state elections to even think about calling any of them, with the possible exception of Colorado, where early voting in 2014 mid-term and 2016 presidential were/are norms; i.e. Pennsylvania is not an early voting state, and Philly with the largest urban city population in the state is typically late in making its contribution.

Sofa Spud wrote: »

Interesting site but needs a pinch of salt - PA is not an early voting state

Indeed!

Sofa Spud wrote: »

Part of me just wants to switch off until about 4am.

This would be a very smart move indeed given the nature of US presidential elections.