Tuesday, December 29, 2009

Bad behavior from 3ie

This is from my holiday email from 3ie, an organization (that I am usually quite happy with) that promotes rigorous evaluation and evidence-based policy in the developing world. The figure in the graph embodies one of the tricks that Huff complains about in his classic book How to Lie with Statistics, which is shaving the vertical access so as to visually overstate the amount of change in the dependent variable.


No comments: