SAMPLING

by

Adam J. McKee, Ph.D.

January, 2004

Populations and Samples

                A population is the group of people or things that a researcher really wants to know something about.  A population might be the members of a class, all the CJ students at USM, all the people in a city, all the people in a state, all the people in the US, or all the people in the world.  Most of the time, CJ researchers want to generalize findings to everyone.  We do not want to just say that GRE scores are related to graduate success in the USM CJ department, but that scores are related to success in CJ in general.

                Rarely can we measure what we want to study in every single person in the population.  This is a census, and only the federal government can afford to even try it.  CJ researchers most always resort to using samples of the population they want to study.  When you go to your doctor, he does not take out all of your blood to check your blood sugar.  He takes only a sample.  The idea is that the sample will be close enough to the rest of your blood to gather the needed information.  A sample is a subset of a population.  We use the sample to infer information about the population.  A person or thing in a sample is called an element.

                When we take a sample, we want it to be as close to the population as possible.  The best way we have found to do this is probability sampling.  Elements must have a known, equal and independent chance of being selected.  So that each element of a population has an equal chance of being selected, we use random selection.  The use of random selection lets us (1) control for bias, and (2) use probability theory in our analysis.  Probability theory is the basis of statistical inference.  It lets us make statements about the level of error that is likely in generalizing our findings in a sample to the population.  There are several types of probability sample.

Simple Random Sample

                The simple random sample is taken by selecting elements randomly from a list of all members of a population.  A simple random sample of CJ students at USM is relatively easy to take.  We print out a list of all registered students, number each, and randomly select them using a table of random numbers or a computer. 

Systematic Random Sample

                This type of sampling is usually used with large population sizes where simple random sampling takes up too much time.  In this type of sampling, only the first element is randomly selected.  Then, every second, third, fourth, fifth, tenth, etc. element is included in the sample.  For this method to work, the list cannot be ordered in any meaningful way.  If offenders were classified by offense, for example, we could not use that list for a systematic random sample.               

 Multistage Cluster Samples

                This type of sampling is used where there is no population list available, such as the entire population of the US.  Here, clusters are randomly selected, and individual elements are then randomly selected from these clusters.  The most common type of cluster is the census track. 

Weighted Sample

                When we want to study the effects of specific variables that are not equally distributed throughout our population, we can "oversample" a particular group.  We increase the probability of selecting persons from the smaller group.  If we want to study the elderly, ethnic minorities, or convicted felons, we can take a weighted sample to assure that these are adequately represented in the sample.  If we wish to make statistical inferences about such a sample, we must consider these weights in our analysis. 

 

Nonprobability Sampling Techniques

                These techniques are generally used when random sampling is not available because there is no way to list the population.  If we wanted to study the motivations of crack dealers, we obviously could not get a list of all crack dealers, nor could we draw a random national sample and only consider crack dealers. This would be too expensive because crack dealers are relatively rare. 

Quota Samples

                In this type of sample, the researcher decides what proportions of particular types of subjects are needed in the study.  Elements of the population are gathered until each of the quotas is filled.  For example, a researcher interested in the effects of race on opinions about the police might decide that 33% of the sample should consist of whites, 33% should consist of African-Americans and the remaining 33% should be Hispanic. 

Purposive (Judgment) Sample

                This type of sample is drawn so that elements have a specific attribute, such as being a rapist.  It is often called a judgment sample because the researcher must make a judgment as to whether the subject fits the criteria for selection.  As with other nonprobability samples, it is dangerous to generalize findings from such a sample. 

Availability Sample

                Elements are selected for this kind of sample simply because they are there.  This type of sample is very common, even if very bad.  Using a class of students is the most common example of this type of sample. 

Theory v. Reality of Sampling

                In reality, criminal justice researchers are often forced to use less desirable sample methods. Often, the variable under consideration does not lend itself to random sampling. Specific criminal offenders are examples of this.   This does not necessarily make a study bad or unpublishable.  Use the best method within the reach of the human and financial resources that you have.  Simply put, random selection is rarely ever possible.  Just get as close as you can.  However, this is a different issue from random assignment.  Elements must be placed into groups randomly whenever possible.

This page available at: