Content area
Full Text
Sampling is defined by Merriam-Webster's dictionary as "the act, process, or technique of selecting a representative part of a population for the purpose of determining parameters or characteristics of the whole population" (n.d.). When conducting research with people (adults or children) who are visually impaired, a low-incidence disability, sampling is a major issue. Inferential statistics are based on probabilities that are more accurately predictive with higher sample sizes. The more participants in a study, the more protected the results of the study are from the influences of random error for which a researcher cannot control. When researchers use the expression "controlling for random error" in their articles, they are talking about limiting the impact that factors unrelated to the study have on what they are seeking to measure through their research.In the field of visual impairment, researchers are often forced to use nonrandom sampling methods such as convenience sampling, in which individuals who fit the criteria of a study are identified in any way possible, or snowball sampling, in which researchers ask the participants they have identified to tell their friends and acquaintances about the study. These methods might help researchers obtain the number of participants they desire, but the way the participants are gathered can easily influence the results by introducing unexpected or uncontrolled factors. In both convenience and snowball sampling, all of the resultant participants will generally be from the same geographical area. They may also have similar socioeconomic statuses or ethnic backgrounds. Any of these factors might have an impact on what the study is investigating. If all the participants are similar on one or more factor, it might skew the results of a study. The best way to reduce the influence of uncontrolled factors is to use random sampling, in which study participants are randomly identified from the population of people who meet the criteria for inclusion in the study. Random sampling is, however, generally far too expensive and cumbersome for researchers to accomplish, especially when studying a population with a low-incidence disability like visual impairment. In the article in this issue entitled “Orientation and Mobility Skills and Outcome Expectations as Predictors of Employment for Young Adults with Visual Impairments,” researcher Jennifer L. Cmar addresses the problem of sampling when studying individuals with visual impairments by using a database of thousands of children and youths from across the United States, the National Longitudinal Transition Study–2 (NLTS2), to boost the number of students included in her study and thereby allow her to examine factors in ways she would not have been able to analyze with a smaller dataset. In addition, the originators of the NLTS2 dataset used complicated sampling methods and often over-sampled certain low-incidence groups such as children who are visually impaired so they would have many representatives of these groups in their dataset. In order to reconcile the fact that the sample was not randomly selected, they also developed a set of weighting factors or “weights” that anyone using their data can use in order to apply the sample in NLTS2 to the entire national population. These weights are numbers that are incorporated into the statis tical analysis of the NLTS2 data so that the end results accommodate the complicated sampling methods for the selected sample. By using the weights provided by the researchers who went through the time and effort to collect the information included in the large-scale NLTS2 database, Ms. Cmar was able to directly tie her results to the national population of children and youths with visual impairments. By doing so, Ms. Cmar was able to expand the potential impact of her study questions, because her results are able to be more closely linked to the national population of children and youths with visual impairments instead of there being questions about whether sampling constraints limited the degree to which her results could be generalized.