Revolutions

Public Opinion and Supreme Court Justices

Andrew Gelman links to an interesting paper investigating the link between public opinion and the selection and confirmation of Supreme Court justices. It addresses questions like, "Should Bork blame public opinion for his failure to be confirmed?"; "Did public opinion seal Thomas's confirmation?" and, "Could Harriet Meiers have been confirmed?" 

The answers to these questions lie in an analysis of public opinion data compared to confirmation votes. There is quite a lot of variability from state-to-state in opinions of potential Justices, as displayed in this chart:

The charting and analysis is all done in R (using the lme4 package).

Statistical Modeling, Causal Inference and Social Science: The Next Supreme Court Justice

 

Kentucky Derby bump-chart

Last week I opined that it would be nice to see a chart of the Kentucky Derby done in R. Little did I know the challenge had already been answered! (Thanks for letting me know in the comments.)

I'd heard about Mine That Bird's come-from-behind performance, but hadn't seen the race myself. You can really see him move ahead at the 3/4 mile mark!

Visit Learning R to see how this chart was created using the ggplot2 package.

Learning R: ggplot2 Bump Chart

Visualizing a horse-race

A reader on the Statistical Modeling blog has come up with an interesting application of ggplot2 in R: visualizing the placements over time of horses in a horse-race (click to enlarge):

The code for the chart is available at the link below. This is simulated race data, but it would be cool to see this chart for the recent upset in the Kentucky Derby.

Statistical Science: A horse-race graph!

Mapping the seven deadly sins

Geographers from Kansas State University have created a map of the spatial distribution of the Seven Deadly Sins across the United States. How? By mapping demographic data related to each of the Sins:

• Greed: Average incomes versus total inhabitants below the poverty line
• Envy: Total number of thefts (robbery, burglary, larceny, and stolen cars) 
• Wrath: Total number of violent crimes (murder, assault and rape) per capita 
• Lust: Sexually transmitted diseases per capita 
• Gluttony: Number of fast-foot restaurants per capita  
• Sloth: Expenditures on arts, entertainment and recreation versus rate of employment 
• Pride: An average of the six other sins 

The calculation for Pride seems a bit unimaginative to me: they justify it as Pride being the root of all sins (really?) but perhaps something like press releases issued per capita, or expenditure on sports teams might have been more interesting. Here are the maps for Wrath and Gluttony (click to enlarge):

You can see slideshows of all the other sins in the Las Vegas Sun article. The images above are clearly showing scaled data: I can't see those Gluttony clusters in Texas, Appalachia and the Carolina coast in the slideshow of the raw data. It would be interesting to know how the scaling was done, but I can't find any references to the "GiZScore" they appear to be using. Nonetheless, it's a great application linking geographic demographic data to what might normally be considered "unmeasurable" quantities.

Las Vegas Sun: One Nation, Seven Sins

What a real flu pandemic looks like

Junk Charts has a neat example of what we would today call a Trellis or Lattice chart. Found in an archive of photographs from the National Museum of Health and Medicine in Washington DC, the chart depicts deaths from influenza in New York City from 1889-1919 (click to enlarge).

 

Deaths from Influenza, lobar, and bronchopneumonia and acute bronchitis, New York City, 1889 - 1919. (c) Otis Historical Archives. Reproduced under a Creative Commons license.

I'm astounded by this chart for two reasons. Firstly, it's not clear when this chart was created, but it's certainly the first example of a small-multiples time series that I'm aware of. Secondly, check out the panels for 1918 and 1919 where the scale of the pandemic was such that the chart literally breaks through the axis and leaps into the upper panels. It's not a technique that would work in general, but here it really makes those pandemic years stand out in context.

There are a few other interesting charts in the archive, too.

Junk Charts: Sore-thumb graphics 

R for Econometricians

Political science researcher and blogger Drew Conway points me to a nice little document aimed at introducing R to econometricians. In 50 pages, Econometrics in R by Grant Farnsworth takes the budding R user through the basics of the R language and works through various econometric analyses, including cross-sectional regression, "special" regression types such as quantile regression, time series models and more. It's a relatively recent document [*], and so takes advantage of some of the more cutting-edge R packages, including the Rmetrics suite, ggplot2, and zoo. 

Says Drew:

I came across this resource while researching a problem I am working on, and found it very useful so I thought I share it with the group. This may be more useful for those of using R to perform analysis in the social sciences, rather than strictly as scripting language, but all users will certainly want to keep it as a reference.

You can download the PDF (and the LaTeX source) from the Contributed section of your local CRAN server.

[* Update: Apparently it's been on CRAN a few years, but updated recently. Thanks to reader DE for the correction.]

Grant Farnsworth: Econometrics in R

Visual Numerics acquired by Rogue Wave

Visual Numerics, the company that produces the IMSL mathematics and statistics libraries, has been acquired by Rogue Wave Software. It's been a long time since I used IMSL (given the high-quality free alternatives now available), but the change might impact anyone who does.

DSC 2009: Early registration extended to May 15

It has just been announced that early registration for the Directions in Statistical Computing conference in Copenhagen has been extended to May 15, so it's not too late to get in for discounted rates (credit-card payments now available, too). Billed as "A Meeting on the Future of Statistical Computing", DSC is the longest-running conference to focus on issues central to R, and was the venue for the first joint meeting of the R-core group. Looks like there's a great program lined up with 25 presentations. I'm registered -- see you there!

Conference website: DSC 2009

Careful with the S-word

Market researcher Tom Ewing offers some advice that applies equally well to statisticians -- be careful when you use the word "significant" in its technical sense. Depending on the audience, it could lead to misunderstandings:

Non-researchers tend to misread “significant” as “important” or simply “big”. Which isn’t the case - it can be trivial or small, it’s just unlikely to be fluke or coincidence.

Researchers tend to read “significant” as “interesting”. Which isn’t the case either - even big results can be utterly banal, especially if they simply confirm something you could have guessed, or if they repeat information you already have.

It's good advice in general, but with regard to the latter point we are given the following example:

For example, suppose we give 1,000 people an IQ test, and we ask if there is a significant difference between male and female scores. The mean score for males is 98 and the mean score for females is 100. We use an independent groups t-test and find that the difference is significant at the .001 level. The big question is, "So what?". The difference between 98 and 100 on an IQ test is a very small difference...so small, in fact, that its not even important.

Then why did the t-statistic come out significant? Because there was a large sample size. When you have a large sample size, very small differences will be detected as significant. This means that you are very sure that the difference is real (i.e., it didn't happen by fluke). It doesn't mean that the difference is large or important. If we had only given the IQ test to 25 people instead of 1,000, the two-point difference between males and females would not have been significant.

Personally, I'm not so sure I'd dismiss that significant 2-point difference so lightly. 2 points may not be a meaningful difference in terms of IQ tests, but I'm immediately led to wonder why a significant difference was observed at all. Was there a problem with the sampling, that led to the men and women in the test being different in some way? Was there some kind of problem with the test, that favored women over men? If you get a significant result you don't expect, it's well worth investigating why -- you may find a surprising, and dare I say, significant, problem with the way the experiment was conducted.

Blackbeard Blog: The Significance Problem (via @russhmeyer)

Fisher's tea lady has some competition

A double-blind study investigates whether people can distinguish dog food from pâté. Oddly, although the sample identified by 72% of subjects as the worst-tasting was indeed dog food, the subjects were were no better than random at identifying the sample that was actually dog food.

Marginal Revolution: Can people distinguish pâté from dog food?