ABSTRACT
One operationally convenient method of selecting a one in k sample from a list of units is to select first a random number between 1 and k and then select the unit with this serial number and every kth unit afterwards: thus, to take a 5 per cent sample of households during a population census, one would first choose a random number between 1 and 20 (here the sampling fraction is 0.05, so k = 1/0.05 = 20); if the random number (also called the random start) is 12, then households numbered 12, 32, 52, 72, 92, 112, and so on will constitute a 5 per cent systematic sample. This procedure is known as systematic sampling and ensures that each universe unit has the same chance of being included in the sample: the constant k is known as the sampling interoal and is generally taken as the integer nearest to N /n, the inverse of the sampling fraction. This method has several advantages. First, it is operationally convenient, especially when, as in multi-stage sample designs, information on the lower stage units, especially the ultimate and the penultimate stages (such as households or families or farms) is not available at the central office and the enumerators have to list these units and to draw samples from them. Second, N, the total number of universe units, need not be known beforehand and a systematic sample may be selected along with the listing of the universe units or with a census, if the sampling fraction is fixed beforehand. Third, a systematic sample is spread out more evenly over the universe, so that it is likely to produce a sample that is more representative and more efficient than a simple random sample.
