Simple Random Sampling

• 1). Make a list of every subject in the population. For example, if you want to survey all the students in a large high school, get a list of all students.
  
  
• 2). Decide on your sample size. If you will be testing hypotheses, the sample size will be based on a power analysis, which will tell you how many subjects you need to have. Power analysis can be done in many statistical programs; it also can be done by using reference books such as Cohen's book on power analysis. If you are testing proportions, for example in an opinion poll, then the sample size will be based on how large a confidence interval you are willing to tolerate.
  
  
• 3). Assign a number to each subject in the population. It does not matter how this is done, as long as each number is distinct.
  
  
• 4). Pick random numbers, either from a table of random numbers or from a pseudo-random number generator on a computer, until you have as many numbers as you needed in step 2.
  
  
• 5). Survey the subjects who got those numbers.
  
  

Cluster Random Sampling

• 1). Divide your population into clusters.
  
  
• 2). Using a random number table or pseudo-random number generator, randomly pick clusters from the list of clusters.
  
  
• 3). Either sample the entire population of each cluster (single-stage random clustering) or do simple random sampling in each cluster. The choice depends on how many clusters you select and how many subjects are in each cluster. This, in turn, depends on what natural clusters exist and the relative ease of sampling from many clusters versus only a few.