Dear Brad, Kiross and Kennedy,
Many many thanks for your help regarding this matter. The programme is for CHW in Angola, and we don’t have readily available experts on sampling to consult here.
I thought it was important to have data at county-level, due to the variance of the prevalence of the indicators (although I do not know these values at county or province level- only have data for regional and national levels), and I thought this would be needed for monitoring and evaluation purposes of the programme.
For the sample size calculation, I used the calculator for proportions of Openepi, using the indicator: 11% (estimate for rural areas) as the proportion of children under 6 months exclusively breastfed
http://www.openepi.com/SampleSize/SSPropor.htm
For a finite population of 2700 children under 6 months (calculated as 2% of the study population of 135,000 people), 5% of absolute precision, 95% confidence level, this gave me a sample size of 143, which I then multiplied by a design factor of 2 and added 10% to account for non-response, giving me a sample size of 315 children under 6 months. Converting it to people and families, gives me around 2,250 families. We have the list of households for each community/microarea to do a two-stage cluster sampling. With a cluster size of 20 households, was planning on randomly selecting 7 communities/microareas from each county, which would lead to 140 households per county.
So this would not be a sufficient sample to give county-level data? I would really need to have 18 "independent surveys", one for each county with the sample of 315 children under 6 months/2250 households? Or at the baseline level this is not of crucial importance, since we will then have monthly monitoring data at county level?
We also do not have the resources to do a sample size totalling 40,500 households (2250 per county). For the livelihood approach for stratification, we don’t have information at county-level.
Again, my many thanks.
Jordana