Saturday March 24, 2012
Election analysis, QLD style, from Bill Ludwig (in Fairfax media):
Labor factional heavyweight Bill Ludwig blamed the men of Queensland for Ms Bligh’s loss and warned Ms Gillard would similarly struggle.
”The blokes here just don’t like women, especially women in charge,” Mr Ludwig said. ”The federal election is a long way off, it’s different circumstances, but I don’t think the blokes like Julia, either. The men of Queensland are just very negative towards women.”
Comments Off on “the men of Queensland are just very negative towards women”
Tuesday March 20, 2012
I just found this in the documentation for the 1986 American National Election study (I’m one of the PIs of the 2012 ANES).
The current policy of the National Election Studies is not to include in
public use data sets special analysis weights designed to compensate for
nonresponse or to post-stratify the sample to known population distribution
controls. Analysts interested in developing their own nonresponse or
post-stratification adjustment factors must request access to the necessary
sample control data from the NES Board.
“Best of luck, Love, Warren…”
Comments Off on non-response was ignorable in 1986?
Thursday March 8, 2012
The ALP is at 3.65 at sportsbet; the Coalition is at 1.37 at Centrebet. The cross-book overround there is a mere 0.39%.
Do ya best.
Comments Off on flirting with an underround
Wednesday March 7, 2012
Graph de jour: Romney and Santorum shares of the primary vote cast yesterday in Ohio’s Republican primary, plotted against Obama vote share in 2008. Units of analysis are OH’s 88 counties. Size of the plotting symbol is proportionate to the number of primary votes cast. Lines are smoothing splines in the mgcv package (fits are generated with weights equal to the square root of the number of primary votes). Data from OH SOS.
As the county shades conservative (smaller Obama vote in 2008), Santorum does better. The more the county supported Obama over McCain in 2008, Romney support tends to increase. This is hardly dispositive, but does count as (yet more) evidence of how Santorum will struggle to win the “median voter” in a key state like OH if he were to be the Republican nominee for the November general election.
Fun fact: last person to win the presidency without winning OH: JFK, 1960.
Some other looks at these data. Romney primary vote share (%) against Santorum (%), by county, plotting symbol indicating number of votes cast. Below the diagonal, Santorum beats Romney; above the diagonal, Romney beats Santorum. Romney wins bigger places (urban), more Dems and Indeps, votes that the Republican nominee would need in the general.
A different look, but the same idea: Romney/Santorum margin (percentage points), by number of votes cast in the Republican primary (log scale):
Comments Off on Romney, Santorum Ohio split, general election portent?
Tuesday March 6, 2012
Awful. So we get some nicer planes on UAL from CO (738s, in seat-power in Economy Plus). But the UAL web site looked at least half-decent.
My quick sense is that they’ve imported a lot of elements from the CO site, and it looks awful, IMHO, at least compared to the old UAL site.
How many different typefaces have been jammed onto this page? And images? Even the word “United” appears in FOUR different typefaces.
Screenshot below (click to open up thumbnail), with my personal details blocked out:
Monday March 5, 2012
A good news, bad news thing?
The good news: my 2009 book Bayesian Analysis for the Social Sciences is enjoying some use out there in Stats-land.
Bad news: eagle-eyed readers are [still!] finding typos and one or two things that go beyond mere typos.
Case in point: my thanks to Miaomiao Dong for pointing out that Proposition 2.8 (page 93) contains a false claim, as does the corresponding part of the “proof” (!) at C.10 (p524). Yikes. It happens, I’m sorry, I wrote that bit too quickly. It is a shame when you make an error like that when stating a “Proposition”.
Screen-shot from the relevant part of the errata appears below (click to open PNG in a new window).
Comments Off on typos in BASS, good news/bad news
Saturday March 3, 2012
Another interesting result I mined out of the YouGovUSA/Polimetrix poll (n=1,000, Feb 18-19, 2012).
Respondents were asked who they thought win a general election contest between (a) Barack Obama and Mitt Romney, and (b) Barack Obama and Rick Santorum. In addition to just looking at the marginals, I tabbed the results to these items by who the respondent said they were supporting the Republican primary:
In both cases a majority of respondents think that Obama will win. Fair enough. But what is interesting is the way this breaks out conditional on who the respondent is supporting in the Republican primary.
If Romney is the (hypothetical) Republican candidate, we get thumping majorities saying Romney beats Obama, among Romney, Santorum and Gingrich supporters. Paul supporters want Paul to somehow run (and win), and so the “someone else” category bumps up there.
But look at what happens when Romney supporters are asked about a Obama vs Santorum general election. A majority (56%) of them say that Obama beats Santorum, which, by the way, is right in line with the general population estimate (60% say Obama beats Santorum).
The result is based on not much data: total n is 1,000, and there are only 63 Romney supporters in the (unweighted); in a gen pop survey, most respondents get filtered away from the Republican primary preference question, since its pretty clear that they are not likely to vote in the Republican primaries/caucuses.
I’ll try to pool some other polls on this to see if this result holds up with more data.
Comments Off on 56-37: Romney supporters think Santorum would lose to Obama