jackman.stanford.edu/blog
• Bayesian Analysis for the Social Sciences Wiley; Amazon; errata as of 5/23/13

• 113th U.S. Senate
• ideal point estimates pdf csv 3/11/14
• scatterplot against 2012 Obama vote share pdf
• roll call object: RData
• 113th U.S. House
• ideal point estimates pdf csv 3/12/14
• scatterplot vs Obama vote share pdf svg
• roll call object: RData

## Tuesday February 20, 2007

Filed under: Australian Politics — jackman @ 12:02 pm

Malcolm MacKerras is always entertaining; it is always as much about him, Malcolm, as it is about the substance of the matter. Hence, the “MacKerrras pendulum” (i.e., sort House of Representatives by two-party preferred vote share, which makes a nice picture with 150 Commonwealth electoral divisions). And this recent effort in Crikey:

Let me be the first pundit to predict that the Liberal Party will lose this year’s general election as a whole, and Bennelong in particular.

Not just the first para, or first sentence is self-referential, but the 2nd word…

Anyway, Malcolm goes on to reason that if the national tide is running against the Liberals, then Bennelong voters will move to dump Howard (which strikes me as a tautology, but we’ll return to that in a moment). Malcolm imagines (hears?) Bennelong voters reasoning as follows:

The defeat of the Howard government will be generally predicted. In that circumstance what are Bennelong electors to do? They could say to themselves: “Let us return Howard in Bennelong to show our gratitude to him. Of course, we know he will, upon the defeat of his government, immediately resign his seat and cause a by-election which Labor will win.”

Alternatively, they could (more sensibly) say to themselves: “We know the Howard Government will be defeated. Why do we not kick Howard out of Bennelong also? That would save the cost of a by-election. Let us save that cost and save the Liberal Party the embarrassment of that by-election loss”.

That last sentence is wonderful. Apparently Malcolm imagines Bennelong voters preferring to humiliate Australia’s 2nd longest serving Prime Minister (i.e., Howard loses his own seat), so as to avoid the “embarrassment” to the Liberal Party if they were to lose any subsequent by-election…

There is another possibility, and that is that Bennelong voters behave in accordance with Malcolm’s much-vaunted model of uniform swing, and turf Howard out along with a bunch of other Liberal incumbents. [Incidentally: it might be interesting to look up the margins carried by incumbent prime ministers into Federal elections (and opposition leaders, for that matter): I'm guessing that Howard's margin in Bennelong, 54.1% 2PP (a "notional" or estimated 2PP margin, after some minor redistricting since the 2004 election result of 54.3% 2PP) is not the smallest ever carried by an incumbent prime minister.]

Moreover, Malcolm’s prediction is a difficult prediction to test. Suppose Howard loses Bennelong and the Libs lose the election. Is this because Bennelong voters engaged in the fanciful kind of reasoning Malcolm suggests? Who knows? We might see if the (presumed, pro-ALP) swing in Bennelong is bigger than the swing in similar divisions, but this could happen for any number of reasons. In 2004, Bennelong and neighboring seats on Sydney’s North Shore swung against the government way more than the NSW average; see here [digression: it is tough to know what to make of this re 2007 -- are these seats trending ALP faster than, say, the NSW average, or will regression-to-the-mean kick in and mean that those seats swing less than they did in 2004?].

On the other hand, Malcolm could be wrong in at least 3 ways: the Libs win the election, or Howard wins Bennelong irrespective of the election outcome, so Mumble is right, Malcolm is sticking his head up here (see also Bryan Palmer’s take), and he deserves kudos for that. “Kudos…”

Finally, there’s a bit more from Malcolm I can’t let pass without comment…

If Howard had any sense he would have resigned the office of Prime Minister last year – as I advised him to do. He could then have retired from his seat this year, following the example of John Anderson.

The self-aggrandizing aside, Malcolm is right there. Howard is in a bit of a pickle, with his legacy to history at stake. Howard has certainly left it too late to ride off into the sunset, with the next election coming up this calendar year (or constitutionally/technically, Jan 19 2008 at the very latest). Rudd is not showing the same intemperateness that seemed to be Latham’s undoing. Howard wheeled in Latham, and may yet do the same to Rudd; witness Howard’s “each-way Kevin” line on Rudd’s Iraq pronouncements, and the claims that Rudd is sabotaging the water deal with the states. More gazing at polls and/or betting markets for a while yet…

## Friday February 16, 2007

Filed under: computing,statistics — jackman @ 11:01 am

One of my students, Sasha Goodman, just cooked up the following web site for finding help on topics in R. Its like a front-end to google, but with the searches coming back contextualized to the “domain of R”. Very nice, very nifty. I started using it immediately and loved it. Sash’s been building on to it in real time since he told me about this last night…so its getting even better…

### imposing normalizing constraints in BUGS, IRT models

Filed under: computing,statistics — jackman @ 10:18 am

Lily Chen writes

Hi all,

I have a IRT model as followings:

model{
for (i in 1:NE){
theta[i] ~ dnorm(0,1)
for (j in 1:NI){
p[i,j] <- phi(a[j]*theta[i] - b[j])
y[i,j] ~ dbern(p[i,j])
}
}
for (i in 1:NI){
a[i] ~ dnorm(mua,siga) I(0,)
b[i] ~ dnorm(mub,sigb)
}

mua ~ dnorm(0,.0001)
mub ~ dnorm(0,.0001)
siga ~ dchisqr(.5)
sigb ~ dchisqr(.5)
}


The thetas’ distribution is distributed as N(0,1).
My question is why the variance of the point estimates of thetas is not close to 1?
The variance of the estimates of thetas is like 0.3.
How can I make the estimates of thetas to have a variance of 1 as I speficifed in the model?
Thanks very much for any hint.

## Tuesday February 13, 2007

Filed under: Australian Politics — jackman @ 9:23 am

A learned colleague noticed John Howard’s unusually strident remarks re Barak Obama and Iraq over the weekend (which did get a run here over the weekend in print/TV news). My colleague remarked:

Howard seems to have lost it! Perhaps it was the shock caused by the
cricket
.

To which I replied:

more like the shock caused by months of opinion polls showing him losing the next election by 4-8 percentage points… he [Howard] had a horrible week in parliament — the 1st week of the sitting year (an election year) — had to go back into the House one evening to correct he an answer given in Question Time on global warming (paraphrase: “yes, I do believe there is a connection between greenhouse emissions and global warming, contrary to what I may have said in Question Time…”). So say what you will about this Obama outburst, but this coming week in Australian politics won’t be about the environment…

And the fun continues… with the debate back on a topic Howard would prefer, but a nice parry from Rudd…

## Friday February 9, 2007

Filed under: computing,statistics — jackman @ 3:08 pm

the Pipes service looks very interesting. well worth a look for those of us who collect/siphon/analyze/publish data for a living… (brought to you from Gruber, from O’Reilly), maybe a hint of what lies in store. That is, once quantitative data starts getting routinely tagged with XML, we’re off to the races…???

## Thursday February 8, 2007

Filed under: Australian Politics — jackman @ 3:31 pm

Peter Hartcher in the SMH reports on the 56-44 Newspoll recently published in Australia. I blogged on the implausibility of these types of numbers before Christmas, when AC Nielsen produced the same 56-44 numbers; see also Andrew Leigh’s recent blog post, quoted by Hartcher. Hartcher’s piece also references some earlier (1993!) work I did with Gary Marks, trying to forecast Australian federal elections using marco-economic conditions.

That work is worth an update, and it might be a useful augmentation to the Leigh/Wolfers argument that “if you want to predict elections, look at the betting markets” (my paraphrase). Of course, the argument re the accuracy of the betting markets is true, but it is worth highlighting the difference between prediction and explanation: the betting prices may (and generally do) predict well, but don’t really move us any closer to understanding why elections turn out the way they do. Put a little differently, what information are players in the betting markets using so as to form forecasts about election outcomes? Macro-economic conditions and published poll results would strike me as two freely available sources of information relevant to forming those beliefs.

### the AJPS goes Interstate

Filed under: type — jackman @ 3:17 pm

in case you were wondering, the typeface on the front and back cover of the rebranded American Journal of Political Science is the very popular Interstate.

### WordPress 2.1 and Ecto not getting along

Filed under: computing,general — jackman @ 9:29 am

I upgraded my blog to WP2.1 and discovered that Ecto could no longer update its post list (would hang at “retrieving categories”). A solution is now circulating.

## Wednesday February 7, 2007

Filed under: general — jackman @ 6:49 pm

From an e-mail that went out from Berkeley early today, has been passed through many hands in the political science world in the last 12 hours…

… I’m very sorry to have to report to you that Nelson Polsby passed away last night (Tuesday, Feb 6). Nelson died peacefully, at home, in his favorite chair with his wife, Linda and daughter Emily Polsby near his side.

I am trying to help Linda get the word out to colleagues who would want to know about this sad development, and wanted to let you know so that you might help spread the word in the political science community and in the Washington community as well. This is a great loss for Berkeley, for the University of California, and for those who study and those who love American politics.

### Know Your Product (Eric Gill bio)

Filed under: type — jackman @ 6:14 pm

Simon, Per your obsession. Best, Eric

Eric Lawrence alerted me to an Atlantic piece on some books on type… some old (a 1989 book on Eric Gill), some new