Monthly Archives: April 2013

Journal club: Effects of Newhints home-visits intervention

This week’s paper Effect of the Newhints home-visits intervention on neonatal mortality rate and care practices in Ghana: a cluster randomised controlled trial, by Kirkwood et al. Cluster randomized trials and meta-analysis, a good combination to put new results in the context of the old.

meta-analysis

Comments Off on Journal club: Effects of Newhints home-visits intervention

Filed under global health

hello, world of statistical graphics in IPython notebook

A few months ago, I had great success invoking the internet to come up with the “hello, world” of statistical graphics.

There are some exciting new developments in javascript-based plotting, and this graphic is just the thing to compare them. D3js has conquered the world in recent years, and is something that my colleagues are starting to think they need to know. Meanwhile, one of the d3js instigators has unveiled the next in his series of revolutions in data visualization, Vega. This is still in development, but may be more appropriate than d3js for routine plots. And it was very soon after the Vega specification and runtime appeared that a python package for it was also released.

Here is an IPython notebook comparing all of these options. The notebook doesn’t save javascript in a way that redisplays, but if you put it in your own notebook server and execute all the cells you should see something like this:

vincent_vega

p.s. google vincent vega to learn the pop culture joke behind this strangely named python package.

2 Comments

Filed under software engineering

Journal Club: Secret Paper

This week in journal club we are reading something that I’m not going to name, because it says “do not cite or distribute without permission” on the top of the paper. This secret paper sounds interesting, maybe I can tell you about it some day.

Comments Off on Journal Club: Secret Paper

Filed under global health

Methods debates

An interesting exchange has been going on regarding the GBD 2010 estimates of deaths due to HIV, published as part of our GBD work last December, along with mortality estimates for 234 other causes of death. The UNAIDS reference group comment, and my colleagues respond.

Comments Off on Methods debates

Filed under global health

ML in Python: Naive Bayes the hard way

A recent question on the PyMC mailing list inspired me to make a really inefficient version of the Naive Bayes classifier. Enjoy.

Comments Off on ML in Python: Naive Bayes the hard way

Filed under machine learning

The Mastery Table for Planning Classes

I’m quite taken with the Software Carpentry approach to teaching scientists computer skills, especially since I saw it in action in UW a few months ago. One aspect that I’ve been trying out for my own course is the “mastery table” approach that the Software Carpentry Instructor Study Groups use. Here is a mastery table for teaching version control. I have made a few of my own, but I don’t think I said enough for any novice to leave competent, according to my ambitions. I will keep trying.

2 Comments

Filed under education

Journal Club: India’s Janani Suraksha Yojana, an Impact Evaluation

We’ve selected a locally grown paper for discussion in journal club this week, India’s Janani Suraksha Yojana, a conditional cash transfer programme to increase births in health facilities: an impact evaluation, by Lim et al, with a focus on the methods: the paper has “used three analytical approaches (matching, with-versus-without comparison, and differences in differences) to assess the effect of JSY on antenatal care, in-facility births, and perinatal, neonatal, and maternal deaths.”

1 Comment

Filed under global health