In defining a population for study, such a population must be specific enough to provide readers a clear understanding of the applicability of your study to their particular situation and their understanding of that same population. It therefore becomes important to select the proper method of sampling, the process by which representative individuals are randomly selected to provide insights into the entire population under study. The four methods of doing this include simple random sampling, stratified sampling, cluster sampling and systematic sampling.
Simple random sampling entails defining the population to be studied, determining the percentage of this population to be interviewed or studied, assigning each individual within the population a number and then using arbitrarily selected numbers from a table of numbers, giving each individual an equal chance to be selected for inclusion in the study. In this manner, a sufficiently random sample of the general population becomes representative of the larger whole.
Stratified sampling involves identifying subgroups of the population representative of the percentages of those same subgroups in the general population being studied, or to equal numbers of individuals within different subgroups for the purpose of comparing their responses to those of other subgroups. Again, as with the simple random sampling, once the population is defined and a sample size is determined, all members of the general population are classified within one of the identified subgroups of the population. Again, a random number is selected and each ninth individual, for example, is interviewed or studied.
Cluster sampling randomly selects groups rather than individuals to be included in a study. The procedures for selecting a sample are the same as a simple random sampling, except that we are now selecting random classes of French 101 students out of the overall population of French 101 classes across the Eastern seaboard.
Systematic sampling is largely the same process, except that it involves selecting an individual or cluster at random and then, in accordance with the desired sample size, including every eighth or tenth or 25th person or cluster in the study.
Nonrandom sampling is the alternative used when random sampling is not a viable option. There three primary methods include convenience sampling. This is the practice of grabbing whoever is handy or willing to participate. Purposive sampling involves selecting individuals known to meet certain clear criteria, such as attending a prestigious tractor repair school in Vermont or Maine. The third form of nonrandom sampling, called quota sampling, relies on interviewing or studying members of particular demographic subgroups, such as working mothers aged 12-18, or 35-49 year old men who regularly attend Star Trek conventions while dressed up like Klingons.
Sampling methods differ somewhat in qualitative studies. These methods include intensity sampling, whereby participants are selected who enable study of differing levels in the subject under study, such as winning poker players and consistently losing poker players. Homogeneous sampling is the selection of people of similar outlooks and/or backgrounds and examines their views as a group. Criterion sampling is apparently the alternative name of choice for purposive sampling when it’s applied in a qualitative study. Essentially it is no different; a set of criteria such as clowns who left the circus to become stop clock salesmen is the basis for individuals being selected for the study. Snowball sampling is a “bring your friends” approach, in which participants identify other people who would also meet the requirements of a particular study. Finally there is the antithesis to this approach, which is called random purposive sampling, in which a select few of a larger qualified group are randomly selected to participate.
Research variables include nominal variables, which classify subjects sharing some common characteristic(s) into two or more categories, such as hat owner or non-hat owner, conservative or liberal, or turtle soup collector or non-turtle soup collector. Ordinal variables rank subjects in numerical order from highest to lowest, such as NBA point guard’s seasonal assist totals, or widest to least wide picture frames. Interval variables have the same characteristics of the first two varieties of variables, but also have equal intervals, so that a test score of 39 differs from a test score of 42 in exactly the same proportion as a score of 57 differs from 60. Ratio variables are similar but start their scale at true zero, assigning it a value – predictably – or zero. Thus my mom’s score of 137 indicating the number of kicks required to break down a door is exactly proportional to the score of 1 kick for my boyfriend (assuming the door and all other conditions were identical).
All of this leads us to the measurement instruments, or as we in the educational field refer to them, “tests.” What is being measured must be unambiguous, the test instructions must be clear and uniformly applied, and the interpretation of the test results must be objective.
A research project, if it to be of any use at all, must accurately measure those variables that form the basis of the research. The criteria must be clear and the researcher’s conclusions must be demonstrable, based on the data collected. The collection of the data must be unbiased, and the researcher must be willing to accept the outcome, even if it disproves their thesis and/or initial constructs.