Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Since the extremely high prevalence of stunting in Northern Nigeria is one of the main concerns that the CDGP aims to address, the power calculation detailed below is based on the measurement of changes in child height-for-age z scores (HAZ). They are measured in standard deviations and calculated as follows:
HAZ=measured height-median height in reference population for same age and sex/standard deviation of reference population
The reference population used here is the 2006 WHO Multi-Centre Growth Reference Study, which combines observations from a range of different ethnic groups from six countries to construct a universal growth standard. A HAZ of 0 means that the child’s height is normal for their age and sex. According to global convention a score of less than -2 indicates stunting and a score of less than -3 indicates severe stunting.
In order to predict the likely size of the effect that the CDGP may have on HAZ, we conducted a review of other impact evaluations of unconditional and conditional cash transfer programs and nutrition education interventions. Based on existing evidence, we have conducted our power calculations with the aim of detecting a 0.2 SD change in HAZ or children in target households in the three years between baseline and endline. This is the expected change between Treatment Group 1 and the control group. Similarly, we aim to detect a 0.2 SD change in HAZ between Treatment Group 1 and Treatment Group 2.
We have estimated the required sample size per cluster assuming 210 clusters. As we have two treatment groups and one control group, this translates to 70 clusters per treatment type. All calculations are based on a power of 0.8 and significance level of 0.05.
With a MDE of .2, and an assumed ICC of .1 (.05), we require 12(8) pregnant women per cluster to be surveyed. This gives a total of 2520 (1680) pregnant women in total to include in the evaluation.
Using estimates derived from the NDHS for Nigeria (2008) we expect to be able to find such numbers of pregnant women at baseline in the surveyed states.