When (if ever) is it reasonable to assume that a sample is not needed because “everyone is the same”—that is, the population is homogeneous? Does this apply to research such as that of Stanley Milgram on obedience to authority? What about investigations of student substance abuse? How about investigations of how people (or their bodies) react to alcohol? What about research on likelihood of voting (the focus of Chapter 9)?
All adult U.S. citizens are required to participate in the decennial census, but some do not. Some social scientists have argued for putting more resources into a large representative sample, so that more resources are available to secure higher rates of response from hard-to-include groups. Do you think that the U.S. Census should shift to a probability-based sampling design? Why or why not?
What increases sampling error in probability-based sampling designs? Stratified rather than simple random sampling? Disproportionate (rather than proportionate) stratified random sampling? Stratified rather than cluster random sampling? Why do researchers select disproportionate (rather than proportionate) stratified samples? Why do they select cluster rather than simple random samples?
What are the advantages and disadvantages of probability-based sampling designs compared with nonprobability-based designs? Could any of the research described in this chapter with a nonprobability-based design have been conducted instead with a probability-based design? What are the difficulties that might have been encountered in an attempt to use random selection? How would you discuss the degree of confidence you can place in the results obtained from research using a nonprobability-based sampling design?