Systematic sampling is popular because it is accurate and relatively easy to use. Systematic samples are taken from sites that are equidistant from each other, either in one or two dimensions, forming a grid. Select the first unit at random. Take the following samples at uniform intervals.
Figure 4a consists of a grid formed when two sets of equidistant parallel lines are intersected at right angles to each other. Figure 4b consists of equidistant, parallel lines, set at 60 degree angles. Triangles are formed by drawing horizontal lines through the intersections.
Figure 4. Grid patterns for systematic sampling.
Although the systematic grid pattern usually allows more accurate sampling, under certain conditions, it is inefficient. Madow (1944) found this to be the case when a fertility gradient existed, either along the rows or columns of a field. For this reason, systematic designs cannot be applied to fields having a slope or drainage problem without first considering the form of population distribution.
The main problem of systematic sampling is how to estimate sample error. There are several approaches:
Assume that the population was in random order before the systematic sample was drawn. Estimate the sample error in the same way as the simple random sample
Block or stratify the sample, assuming that variations among units within a block are sample variations. Estimate the sample error as with the stratified random sample
Take a number of separate, systematic samples, drawn at random from all possible systematic samples of the same type. Calculate the mean and treat as a simple random sample.