Sample size

The sample size needed depends more upon the variability within the population than upon the size of the population. If a probability method of sample selection is used, a statistical method is needed to estimate the size of sample. Two basic formulae are available - one for those studies where an estimate of the mean is the main survey objective and one for those studies where proportions are the main survey measures to be obtained.

If a non probability method of sampling is chosen, the researcher should ensure that a sufficient number of respondent exist in each category of the important variable. There appears to be some agreement that there should be at least 100 in each stratum. For less important variables, somewhere between 30 and 50 would suffice. If the researcher wishes to analyse the results by size of firm (for example) they should ensure that there are at least 100 firms in each size category.

Cost of course, if often a constraint on the sample size finally used.