Protected: Two ways to estimate primary outcomes without polls (transcript)
This content is password protected. To view it please enter your password below: Password:...
Sam Wang in Ohio – today at 3:00pm, Cincinnati
Usually I wait for the data to come to me…but today I am in Cincinnati, Ohio. Come see me!
Today at 3:00pm, I give a public lecture just across the river at Northern Kentucky University. It’s open to the public. I will probably tell the Ohio voters in the audience that if they voted for Kasich, they were effectively helping Trump. That should go over well. I’ll also mention the one strategy Kasich could follow that would slow down the Trumpocalypse.
Details: “Who Needs Pundits and Pollsters? with Dr. Sam Wang” is 3-4 p.m. Wednesday, March 16 at the George and Ellen Rieveschl Digitorium in NKU’s Griffin Hall. The lecture is free and open to the public, but registration is requested. The College of Informatics Tenth Anniversary Distinguished Speaker event is part of the college’s yearlong anniversary celebration.
Topics:
Yes, a vote for Kasich is indeed a vote for Trump, but of course in the absence of data how do we know some fraction of Ohio voters did not intend that very outcome?
If this is no candidate arrives with >1237 delegates, a reasonable question to ask would be what does the probability of delegate switching have to be to give a given candidate the nomination?
More precisely, if we go by Fig 2 in the article and Trump has around 1100 delegates and Cruz and Kasich have xx and yy, what max can the ratio of switching probs have to be to give Cruz the nomination? p_(Trump->Cruz)/p_(Cruz->Tump) = ? (and same ratios for Kasich).
Also, do we have any priors for these switching probabilities?
What’s making your histogram so much different from the figure on the front page of the NYT this morning? http://www.nytimes.com/interactive/2016/03/16/upshot/trump-cruz-kasich-republican-delegate-lead.html
It looks as if the NYT simulator is assuming a constant (adjustable) fraction of popular votes for Trump, Cruz and Kasich across all future primary states, which is obviously not realistic. Sam wasn’t assuming that, was he?
I assume more complexity: +/-10% uncertainty (standard deviation) in the #1-#2 margin in states where there is a poll median. Where there are no state polls, +/-16% uncertainty (again SD) around the national #1-#2 margin. This latter number is a little more than the actual variation in Romney’s margins in 2012.
My new favorite word
Trumpocalypse
Try not to think of the Zombie Trumpocalypse should there be a contested election! Carly Fiorina, Jeb Bush and even Scott Walker would be creeping off the dissection tables lusting for Republican delegate flesh.