My Photo

March 2017

Sun Mon Tue Wed Thu Fri Sat
      1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31  

BlogStats


« Predicting with Expressed Beliefs – a formal approach | Main | Responding to Paul D. Hauck (updated 17feb16) »

07 February 2016

Comments

George Rebane

To start things off Jo Ann and I took independent cuts at the outcome of the New Hampshire primary next Tuesday. A convenient way to express a bunch of MABs whose realized values have to add up to a certain number is to first pick your best guess or most likely (M) values and make them add up to that certain number – here we took that to be 100% of the vote for each party. Then go back and put in the L/H values that you think bracket your best guess in the smallest reasonable range, and finally review your best guess and write in your confidence C values. So here are our [L, H, M, C] tuples.

JAR

Trump 19%, 26%, 21%, 0.6
Cruz 15, 20, 17, 0.2
Rubio 17, 28, 20, 0.3
Kasich 9, 15, 10, 0.3
Bush 5, 10, 6, 0.3
Christie 8, 12, 9, 0.4
Fiorina 8, 12, 10, 0.4
Carson 5, 10, 6, 0.3

Clinton 20, 40, 30, 0.4
Sanders 40, 80, 70, 0.3

GJR

Trump 25%, 38%, 28%, 0.5
Cruz 15, 25, 20, 0.6
Rubio 15, 20, 16, 0.4
Kasich 5, 12, 8, 0.3
Bush 8, 16, 10, 0.5
Christie 6, 12, 8, 0.3
Fiorina 2, 6, 5, 0.3
Carson 4, 7, 5, 0.3

Clinton 25, 50, 35, 0.5
Sanders 50, 75, 65, 0.6

Russ Steele

George, After listening to the Megan Kelly's interviews, this is by best guess.

RWS

Trump 19%, 26%, 21%, 0.7
Cruz 15%, 20%, 17%, 0.3
Rubio 17%, 28%, 20%, 0.2
Kasich 10%, 15%, 11%, 0.5
Bush 4%, 10%, 5%, 0.2
Christie 8%, 12%, 10%, 0.4
Fiorina 8%, 12%, 10%, 0.3
Carson 3%, 7%, 6%, 0.3

Clinton 20%, 40%, 30%, 0.3
Sanders 40%, 70%, 60%, 0.5

George Rebane

RussS 643pm - So noted Russ, good luck ;-)

Russ Steele

How Google Searches Pretty Much Nailed the New Hampshire Primary

Real-time trending search queries foretold the outcome of the election.

Google’s ability to look into the future of political contests just notched another win: New Hampshire.

Searches of presidential candidates conducted by Google users in New Hampshire on Feb. 9 corresponded closely with the voting results of the state’s primary. The top-searched Democratic candidate was Bernie Sanders, who won with 60 percent of the vote in New Hampshire, according to the Associated Press. He got 72 percent of the searches, according to Google, while Hillary Clinton got 28 percent of the queries and 38 percent of the vote.

The top-searched Republican candidate was Donald Trump, who won with 35 percent of the vote. On Google he received 41 percent of the searches an hour before the polls closed, according to the search giant. No. 2 was John Kasich, who got 16 percent of both the vote and the searches. Ted Cruz took third with 12 percent of the vote and 15 percent of the searches. The battle between Jeb Bush and Marco Rubio was close online and in real life. While Bush took fourth place at the polls, winning 11 percent of the vote, online he got just 7 percent of the searches. Meanwhile, Rubio got 10 percent of the searches and only 10.6 percent of the vote.

http://www.bloomberg.com/news/articles/2016-02-11/how-google-searches-pretty-much-nailed-the-new-hampshire-primary

Graphics at the link.

George Rebane

RussS 612pm - That's all fine but 1) you can't use their data before the election, i.e. before they publish it, and 2) Google doesn't say anything about YOUR or MY ability to predict.

Russ Steele

George@11:34AM

Google has a Real Time Reporting API that will let a smartphone user monitor search terms right up to the time a voter pulls the handle.

Check out this chart at 4PM Feb 9th:

https://www.google.com/trends/story/US_cu_4JIPz1IBAAC93M_en

It appears to me that it is possible to monitor the search terms in real time. Set up the search and then use the Real Time Reporting API to monitor the results.

True, hard to predict days in advance, but it looks like it is possible to predict the outcome before the polls close.

George Rebane

RussS 132pm - No doubt you are right. But unless you are trying to win a specific wager while the polls are open, that kind of information is of little use, and most certainly does not reflect on your acumen in predicting. Predictions lose both their utility and panache the closer they are made to their resolving event.

I'll let my 1134pm stand.

The comments to this entry are closed.

Blog powered by Typepad