Systematic Random Sampling
Systematic random sampling techniques are easy for those who have access to a sampling frame. Once the researchers have a list of the population, they will choose every nth (5th, 10th, etc.) participant to obtain the desired sample size. When I was conducting an evaluation at my local county jail, I needed a sample of 500 recently released individuals to serve as a comparison group for a reentry program evaluation. The jail staff identified 5,000 potential comparison group members, so I asked them to select every' 10th person on the list. A potential drawback and threat to randomization is if the sampling frame is listed in some sort of order that could generate sample bias if every' certain number of cases or people are selected. It is recommended that researchers select a random start point on the sampling frame and then proceed with picking every nth case.
Stratified Random Sampling
Stratified random sampling requires that researchers are a bit more knowledgeable about the characteristics of the sampling frame from which the sample is being selected. That is because stratified random sampling involves dividing the population into different strata, often based on demographics such as sex, race, age, or socio-economic status, depending on how important adequate representation of these groups is for the study (Hagan, 2007). From there, researchers select random samples of participants from each stratum. This sampling plan is appropriate anytime it is particularly important for certain segments of the population to be adequately represented. Randomization maximizes, but does not guarantee, a representative sample. Stratification safeguards researchers against accidentally randomly selecting mostly whites for a study when it is essential to hear from various racial groups.
Stratified sampling techniques can be proportionate or disproportionate. Proportionate stratified samples include selection of enough cases or individuals to ensure that each stratified group has the same ratio of representation in the sample as they have in the population. It might be desirable to select a sample that has proportions of representation that do not match the population, and that is when it is appropriate to conduct disproportionate stratified sampling. For disproportionate stratified sampling, we would oversample some groups and undersample others. Researchers who are looking to study the impact of racial discrimination on the lives of residents of a particular county or state would likely want to oversample racial minorities.
Multistage Cluster Sampling
As I noted, stratified sampling requires some knowledge about the individuals or cases that are being stratified, as their memberships in certain groups needs to be apparent to the researchers in the early stages of sampling. Multistage cluster sampling involves division of census tracts, blocks, other areas, or even population units and then taking a probability sample from each cluster (Hagan, 2007; Maxfield & Babbie, 2012). This can be particularly useful when a sampling frame is not available. When studying the impact of police intervention on possible crime displacement, Weisburd et al. (2006) identified both intervention areas and nearby catchment areas that might be impacted by spatial displacement. Data collection involved random selection of street segments within those clusters for social observations and random selection of households in the same clusters for victimization surveys.
The National Crime Victimization Survey utilizes a stratified, multistage cluster sample to select households for their panel study of victimization. The first step of their sampling plan is to stratify by primary sampling units (PSUs), including counties, groups of counties, and large metropolitan areas. All the large PSUs are automatically included in the sample, but the smaller ones are grouped by geographic and demographic characteristics derived from the US census. After randomly selecting some of the smaller PSUs, the researchers divide all the PSU areas into types of housing (such as group quarters, individual houses, etc.) and select clusters of units from those housing types (United States Bureau of Justice Statistics, n.d.).