Systematic Sampling and Choosing a Method

Systematic sampling starts with a list of n items and a chosen interval k. You pick a random starting point among the first k names, then take every kth name after that: start, start plus k, start plus 2k, and so on. From a list of 1,200 names sampled every 30th, you first pick a random number from 1 to 30, say 17, then take names 17, 47, 77, 107, and so on until you run out of list. That gives you about 1,200 divided by 30, or 40 names. It is fast, cheap, and spreads the sample evenly across the list, which often makes it behave like a simple random sample.

The danger is periodicity. If the list has a repeating pattern and your interval k matches that pattern's cycle length, every name you pick shares the same trait. An alternating male, female, male, female list sampled every 2nd name gives you all one gender, not a mix. The fix is to check the list's order before trusting the method, or to choose a k that does not divide evenly into the pattern's cycle.

This is not the same failure as a convenience sample, which grabs whoever is easiest to reach, like the first 40 people who walk by. Making a convenience sample bigger does not fix it, because the bias is in who gets a chance to be chosen, not in how many you choose. Systematic sampling, by contrast, gives every name on the list a fair shot through its random start, so it belongs with simple random, stratified, and cluster sampling as a real probability method. Choosing among them depends on the list: use stratified when you want guaranteed representation of known groups, cluster when travel or cost favors grouping, and systematic when you have an ordered list with no hidden rhythm.