Findings, implications and lessons learned

• The most efficient form of sampling in terms of precision of population estimates is to sample proportional to size. Thus, the 60 dip-tank areas to be sampled would ideally be distributed so that similar proportions in each sub-region are sampled relative to the total number of dip-tank areas in the sub-regions.

• The 600 or so homesteads within the Hhohho Sub-region would then be distributed across selected dip-tank areas proportional to the numbers of homesteads within the areas. This would be preferable to that of choosing a constant number, as was the intention in this case study, for each dip-tank area.
• Finally, it should be noted that different statistical software that can be used to carry out calculations of standard errors of population estimates derived from cluster sampling sometimes result in slight differences in results, and may even differ from those derived here.

More than likely this may be due to slight differences in the construction of the formulae. One should nevertheless verify that this is indeed the reason and not a result of misunderstanding of how the calculations have been done.


Source:Thalile Seipone Sowane