Systematic random sampling involves selecting every nth unit of the target population from a list that is randomly ordered—for example, every 10 th spousal caregiver of breast cancer patients. It is important that the population listed is not ordered in some way to create a bias, for example, in this case, if every 10th caregiver was Caucasian or lived in the southern part of the United States.
Stratified random sampling involves dividing a population into groups or strata and then randomly selecting from that group. The strata might be age, gender, cultural/ ethnic group, level of experience, or level or duration of a disease. Researchers use this technique to ensure that subgroups of a population have an equally likely chance of being included in a sample. For example, in our caregiver example, assume that the population of caregivers is 50% White Caucasian, 30% Black/African American, 10% Hispanic, and 10% Asian, and the goal is to examine if responses to the intervention vary by race/ethnicity. Using stratified random sampling, the population could be divided into racial/ethnic groups, and then individuals would be randomly selected from each of these groups. Using this strategy would guarantee that the sample would include members from all of the race/ethnic groups. A proportional stratified sample is one in which the size of each strata in the sample is proportional to the size of the strata in the population; whereas in a disproportional stratified sample, the size of the strata is not proportional to the actual size in the population.
Again, stratified random sampling should not be confused with stratified random assignment of participants into treatment groups. For example, in the REACH II trial, convenience sampling procedures were used to recruit study participants. However, the randomization scheme was stratified by race/ethnicity to ensure that equal numbers of Black/African Americans, Latino/Hispanics, and Caucasian/ Whites were included in the intervention and enhanced (information only) control conditions.