- My book: Bayesian Analysis for the Social Sciences (Wiley; Amazon).
- 111th U.S. Senate ideal point estimates:
- 111th U.S. House ideal point estimates:
- Next Australian election, Centrebet prob. of ALP win: 0.78 (2010-02-10, time-series)
Some graphs looking at the voting on the Stupak amendment. This roll call sliced up the Democrats pretty nicely. Thumbnails below link to PDFs. Democrats only in the 1st graph, looking at the relationship between the Ayes and Noes and Obama vote share in the representatives’ respective districts.
RSS feed for comments on this post.
Sorry, the comment form is closed at this time.
Powered by WordPress
Bad Behavior has blocked 397 access attempts in the last 7 days.
Thanks!
Comment by kevin — Monday November 9, 2009 @ 6:57 am
Thanks for the nice post.
Which R function/package did you use for the plot?
Comment by Thomas — Monday November 9, 2009 @ 7:46 am
Thomas: This was all done with custom code. I create a plotting region with a dummy call to plot with type=”n”, then start adding element as necessary.
Comment by jackman — Monday November 9, 2009 @ 11:18 pm
Dear Simon, thank you very much for the info.
I was mainly pondering the following two questions:
1.) Did you use some kind of kernel estimator for the line plot in the middle (between the point rugs) and if yes, which one did you use?
2.) Is there a way to quantify the validity of such plots? e.g. employ bootstrapping to get an estimate of the variance for the “ideal point” or does there exist some hypothesis test for such an scenario?
Comment by Thomas — Tuesday November 10, 2009 @ 2:22 am
Thomas:
On (1), the curve, I fit a binomial model (Y/N on ideal points with a probit model, since that is what we use in the model that recovers the ideal points; or logistic regression when the predictors is Obama vote share in the members’ districts) using gam in mgcv, with a smoothing term on the continuous predictor; in these examples I generated over the weekend I always wound up with a 1 degree of freedom fit to the data, but I like to at least explore the possibility of non-linearity.
Take a look at the regression analysis I have of House members’ ideal points as a function of Obama vote share here, or linked at the top of my blog (and updated daily). Those are loess fits with the default 2/3 span (panel.loess in lattice).
On (2) on validity, note that the ideal points are recovered from the votes, so they will always fit well, in some sense.
Questions about validity might be of the sort “are the labels `liberal’ and `conservative’ appropriate?” or, “have we fit enough dimensions to the data”? I’m generally happy with answering “yes” to both of these questions, although the split among the Democrats on the Stupack amendment (with abortion as the driver there) suggests that there could well be a 2nd dimension at work in the data (at least among Democrats).
Is this what you had in mind?
Comment by jackman — Tuesday November 10, 2009 @ 2:36 am
Dear Simon, thanks a lot for the detailed explanation.
I’ll take a closer look on the material you have published and then experimentally apply it to my data.
Keep up the good work!
Comment by Thomas — Tuesday November 10, 2009 @ 3:19 am