Inequality vs Stuff

I went to a talk a few weeks ago by Richard Wilkinson and Kate Pickett, global health researchers who have written a book called The Spirit Level.  They were quick to explain that, while the name makes perfect sense in British English, it has been a source of continuing confusion in American English.  What is a “spirit level”?  It’s a building tool, a type of ruler with little bubbles in it to show when it is parallel to the ground.  Maybe it’s called a carpenter level in the states, or just a level when the context is clear.

I would have called it “Inequality vs Stuff”, or at least that’s my description of the talk:  a vast array of scatterplots showing the relationship between income inequality and different measurements of population health.  Here is one that is typical for their case:

When they told the story, they started with a composite health index scattered against inequality, since that has much less noise, and then use the noisy plots like this one as supporting evidence when they show that the relationship holds for everything.

The slide that stuck with me the most is one that diverged from their story a little:

Not population health this time, but still interesting.  Something to share with your entrepreneur friends.

These plots seem like enough fun that I made my own, based on a question from the question and answer portion of the talk.  I’ve forgotten who, but someone in the audience asked “How is inequality related to total fertility rate?” and the answer from Wilkinson and Pickett was along the lines of “We never thought to check, how do you think it might be related?”

Since I had the data lying around from my attempts to learn about model selection last summer, I made myself the plot.  Turns out there is not much of an association.The only example of a non-association the speakers mentioned was a surprise to them: suicide rates are not correlated with income inequality.

haiti.ushihidi by category

What people are saying now: Water shortage, food shortage, medical equiptment needed.  What they are not saying as much anymore: Food, shelter, search and rescue.

(data details in previous post.)

OR and Crisis Camp

When the earthquake devastated Haiti, Laura McLay asked if OR is helping with the relief efforts.  I’ve been wondering the same thing, and I went to a “Crisis Camp” this weekend to see if there is anywhere I could plug in.

This Crisis Camp business is hard to describe, and I didn’t really know what I was getting into when I showed up, and it seems like most of the other participants didn’t either.  But we all woke up for a 9 AM meeting on Saturday, and we all wanted to do something good for the people of Haiti.

This isn’t exactly something you can make an impact on in a day, and the only tangible result of my work was fixing a typo on a wiki, but I did learn a little bit about what is going on.  One group of geographers did a quick course on Open Street Maps and was able to start helping in an effort to update the maps of Port-au-Prince, tagging blocked roads, collapsed buildings, etc.

I joined group that connected with an ongoing project to find hospitals outside Port-au-Prince and help them share information about available capacity with people who need medical attention.  Like I said, I didn’t manage to help with this in a day, but I did learn about this Sahana project and their success in finding the lat and long of 100 hospitals in Haiti.

Another impressive data sharing tool that I a look at is Ushahidi, which I had heard about before, but never seen in action.  This is a project that has a free SMS gateway for people in Haiti to use to report emergencies or share information.  They translate messages into english and post them on the web with a CC-BY-SA license.  I started looking at them yesterday, and they can be heartwrenching.  Here is the breakdown by category, as of last night:

Does this inspire any operations research solutions?  It makes me think of vehicle routing, if the earthquake damage tags in Open Street Maps show which roads are closed, that is:

I’m not sure if they do.  (Red is map features with the tag earthquake:damage, but those are mostly IDP camps.)

Health Care Reform, Accountability, Disparity

I got some good news for the weekend, an opinion piece that I wrote together with some of the other post-graduate fellows at IHME was published online as a Science e-letter. It is titled U.S. Health Care Reform: The Case for Accountability and it’s about the measuring the outputs, outcomes, and impacts of the reform, whatever shape they end up taking.

The part that I was especially interested in adding to the discussion appears in paragraphs 3 and 4, about what these some of these statistics look like currently:

Disparities in health outcomes in the U.S. are unacceptable. A healthy life expectancy at birth in the U.S. ranks behind 28 other developed countries (1). Sizable groups in the United States have mortality risks resembling those in sub-Saharan Africa (2), including urban blacks between the ages of 15 and 64 living in counties with high homicide rates.

On average, Asian women lived 21 years longer than high-risk urban black males in 2001 (2). Although life expectancy for most American women increased between 1983 and 1999, life expectancy for women in 180 counties in areas such as Appalachia, the Deep South, the southern Midwest, and Texas decreased by 1.3 years (3).

I made some figures to accompany this, which Science didn’t print, so I’ve included them for you here:

Probability of a 45 year-old male dying before age 65, 2001, from Murray et al., Eight Americas: Investigating mortality disparities across races, counties, and race-counties in the United States. PLoS Medicine 2006.

Female life expectancy in US counties, 1961-1999 from Ezzati et al., The reversal of fortunes: Trends in county mortality and cross-county mortality disparities in the United States. PLoS Medicine 2008.

ACO in Python: Shortest Paths in NetworkX

I’m supposed to be doing the final edits on the journal version of this old paper of mine (jointly with Greg Sorkin and David Gamarnik), but I’ve thought of a way to procrastinate.

Instead of checking the proofs that the length of the shortest path in my weigthed width-2 strip is , I’ll make a quick blog post about verifying this claim numerically (in python with networkx). This also gives me a chance to try out the new networkx, which is currently version 1.0rc1. I think it has changed a bit since we last met.

from pylab import *
from networkx import *

G = Graph()
for u, v in grid_2d_graph(100, 2).edges():
    G.add_edge(u, v, weight=rand() < .5)

wt, p = bidirectional_dijkstra(G, (0,0), (99,1))

Multilevel (hierarchical) modeling: what it can and cannot do in Python

I re-read a short paper of Andrew Gelman’s yesterday about multilevel modeling, and thought “That would make a nice example for PyMC”.  The paper is “Multilevel (hierarchical) modeling: what it can and cannot do, and R code for it is on his website.

To make things even easier for a casual blogger like myself, the example from the paper is extended in the “ARM book”, and Whit Armstrong has already implemented several variants from this book in PyMC.

Machine Translation and the Porpoise Corpus

I might have mentioned that I got to do some world traveling for my work recently. Seeing rural Tanzania was an experience that I still don’t really have good words to describe. But this is not a post about that. This is a post about a sticky idea I got stuck on in some science fiction I was reading during my multi-day to and fro travel.

On my around-the-world-in-4.5-days journey, I read the Jewish feminist sci-fi novel He, She, and It by Marge Piercy. It’s got a classic hard AI theme, about a robot that is so, so human… I’d recommend it. But dilemmas of whether a robot can make a minyon in the reform tradition of 2059 has not stuck in my mind the way this one line about whales has:

Paper rejected, Cheer Up with Baby Animals

Too bad for me, my first global health paper will have to be revised and resubmitted. In addition to some more substantive objections, the negative reviewer said “It is unclear what software was used to carry out the Bayesian estimation by MCMC. This is not possible in STATA and would be extremely difficult in the scripting language, Python.” It was difficult in Python! I doubt that any software would make it much easier, though.

To cheer myself up, I’ve been looking into the newest fads in pets: robotic hamsters and teacup pigs.

Post-doc Ops

Would you like to work with me applying computational algorithms to challenges in global health metrics? Then apply for the IHME post-graduate fellowship. Deadline is Feb 15.

(There is also a “pre-graduate” version, for those who have not started graduate school yet.)

Clustering with Shallow Trees

I’m updating my CV, and that reminded me that I meant to promote this cool clustering technique that I was a little bit involved in, Clustering With Shallow Trees.

This goes way back to about half-way through my post-doc at MSR, when statistical physicist Riccardo Zecchina was visiting for a semester, and was teaching me about all of the “intractable” optimization problems that he can solve using his panoply of propagation algorithms. In particular, he was working on algorithms for certain types of steiner tree optimization, and he had discovered that adding an extra constraint on the depth of the tree didn’t make the problem harder. (All variants of the problem he considers are NP-hard, but some are NP-harder than others.)