Study design

Sampling of homesteads was done in two stages of cluster sampling.

In the first stage, a mixture of representative and random sampling of dip-tank areas (primary sampling units - PSUs) was done within each sub-region. The representative nature of the sampling took account of ecological zone, proximity to commercial farms, proximity to urban areas, remoteness and livestock density. Remoteness was taken into account in order not to miss a genetic resource that might be present only in one particular location.

Whilst representative sampling has the attraction that it ensures that all situations can be covered, it has the disadvantage that a standard error cannot be calculated for the estimate of the population total. Furthermore, the estimate of the population total may be biased.

One way round the problem is to define strata within each sub-region based on the criteria mentioned above, and then to sample dip-tanks areas at random from within each stratum. This method was used in this survey.

The numbers of dip-tank areas selected within a sub-region depended on the total number of dip-tank areas and presumed livestock density in that sub-region; in general, the more dip-tank areas the larger was the sample. Thus, there was an attempt to sample from sub-regions approximately, but not entirely, proportional to the number of dip-tank areas.