This "probability sampling" method consists of a number of stages and is designed to retain the advantage of simple random sampling and at the same time cut down the cost of the sample. The method is best explained by taking the bank survey already discussed as an example, and working through the various stages. Suppose you have decided that you need a sample of 5,000 adults selected from all the adults in the UK, but that the expense of running the survey with a simple random sample is too high. Then you could proceed as follows:
Stage 1 Use all the administrative counties of the UK as the sampling units and select a
simple random sample of size 5 from this sampling frame.
Stage 2 Each county will be divided into local authority areas. Use these as the sampling
units for this stage and select a simple random sample of size 10 from each of the 5 counties
chosen in stage 1. You now have 50 local authority areas altogether.
Stage 3 Divide each of the selected local authority areas into postal districts and select
one of these districts randomly from each area. So you now have 50 randomly
selected small regions scattered throughout the country.
Stage 4 Use the electoral rolls or any other appropriate list of all the adults in these
districts as the sampling frame and select a simple random sample of 100 adults
from each district.
If you check back over the stages you will find that you have a multistage sample of total size 5,000 which is divided equally between 50 centres. The 100 persons at each centre will be easy to locate and can probably be interviewed by one or two interviewers. The subdivisions at each stage can be chosen to fit in conveniently with the particular survey that you are running. For instance, a survey on the health of school children could begin with local education authorities in the first stage and finish with individual schools.
... Advantages – the advantages of this method are that at each stage the samples
selected are small and interviews are carried out in 50 small areas instead of in 5,000
scattered locations, thus economising on time and cost. There is no need to have a
sampling frame to cover the whole country. The sample is effectively a simple random
sample.
... Disadvantages – the main disadvantages are the danger of introducing interviewer
bias and of obtaining different levels of accuracy from different areas. The interviewers
must be well chosen and thoroughly trained if these dangers are to be avoided.
0 Comments