Determine the goals of your experiment and the amount of resources, such as time and money, that you have. Goals should include knowing about the population you're studying and what concepts you are trying to measure in your experiment. The amount of resources at your disposal may affect the amount of precision you can strive for in your experiment.
Decide on the level of precision you want in your experiment. Precision refers to how closely the results of your experiment, as measured by your sample, mirror the same measures for the entire population. The difference between the results for your sample and for the entire population is the margin of error (sometimes called a confidence interval). Many studies and experiments decide on a margin of error of plus or minus 3 to 5 percentage points. This means, for example, if your experiment found that 70 percent of the subjects in your sample tested positive for a particular condition, a margin of error of 3 would mean that 67 percent to 73 percent of the population at large would test positive for the same condition. The greater the precision you want, the larger your sample will need to be.
Set a confidence level, or the level of risk that the results for your sample reflect those of the population at large. As with precision, you will see a trade-off between confidence levels and sample sizes. The higher the confidence level desired, the larger your sample will have to be. Many experiments set a confidence level of 95 percent, meaning that if the experimenter drew 100 samples, 95 of them would show the same results. Note that the confidence level should not be confused with the confidence interval, or margin of error, discussed in Step 2.
Consider the degree of variability in the population being studied. Variability refers to how characteristics or attributes of interest to your study are distributed across the population. A homogeneous population has less variability and thus does not require a large sample. A diverse, heterogeneous population, in contrast, has high variability, requiring a larger sample size. For example, a study of school teachers estimating that 80 percent are certified and 20 percent are not would have a variability level of 20 percent, or .20.
Calculate a sample size. Many statistics and research-oriented websites have sample size calculators where you can enter the margin of error (confidence interval), confidence level, population size and variability. After entering these values, the calculator will give you a sample size. The sample size calculator operated by Raosoft (see Resources), for example, calculates that for a population of 1,000, with a margin of error of 5 percent, a 95 percent confidence level, and a variability level of 25 percent, the recommended sample size should be 224.