After last week’s post on random sampling, I received a comment from a friend. She recommended some tools that might help calculate sample size especially when the population is different from the list Dillman offers. It is called Macorr. It has fields for the same variables that Dillman lists AND the population size can vary. Very important since Extension programs that repeat may not have a nice round number in the population.
She also says that you can calculate a random sample in Excel. She has to send me the directions on how to do that. I offer that to those of you who are much more adapt at Excel than I am.
When she sends it to me, I will post it. In the meantime, we wait or use the resources already available.
A comment was made: How important is it to have geographical representation in your random sample?
Theoretically, the random sample allows all individuals in a population to have the same chance of being in the sample. Because of that chance, there is also an excellent likelihood that the geographic representation will be distributed to represent the population. Of course, you have to decide before you sample, to what questions you want answers. If geographic areas may affect the outcome, then I would suggest the following. If you want to make sure that a particular area is represented (i.e., mixed metropolitan and rural areas), you can stratify on the type of representation you want. I’m doing this in an evaluation I will be undertaking this summer and fall. We hypothesize that the metropolitan areas are different from the mixed metropolitan/rural areas and both are different from the rural areas. The evaluation team stratified on the density question and are randomly selecting in the three areas. I’ll let you know how the stratification worked.