Association between the pre–and post–intervention incidence per 1000 in the 40 control and 40 collar villages.

Line: least squares linear fit from regression of the post–intervention incidence against the pre–intervention incidence, weighted by number of children in the denominator (symbol size).