Convenience Sampling Revisited: Embracing Its Limitations Through Thoughtful Study Design

In one of the first Statistical Sidebars I wrote (published in the March–April 2015 issue of the Journal of Visual Impairment & Blindness), readers were introduced to types of sampling. The point was made that when conducting research in the field of blindness, researchers often make use of convenience sampling. A convenience sample occurs when participants who fit a study’s criteria are enrolled in the study, sometimes by simply going to a location that is likely to have a large number of people who are blind (e.g., to a state school for blind students or a convention of a blindness organization). The main drawback of convenience sampling is that the study results lack generalizability due to the bias of the sample. As such, a study that uses convenience sampling is somewhat caught between a single-subject approach (where a treatment or intervention is used with very few participants to focus on the efficacy of that treatment or intervention) and a randomized control group approach (where a sample of participants are randomly chosen from a large population and randomly assigned to treatment and control groups). The use of a larger sample through convenience sampling, rather than utilizing a single-subject approach, allows for slightly greater generalization, but the lack of random selection of participants means that selection bias restricts large-scale generalization of the findings. In the current issue, the author of the article entitled, “Locus of Control in College Students with and Without Visual Impairments, and the Visual Characteristics That Affect It,” used convenience sampling to gather participants who are blind for the study. He put together a sample of 15 participants who are blind. This sample is larger than a singlesubject approach, thus, it has the potential to show that any experimental effects are more generalizable than they would be if the author had studied just one individual. The convenience sampling procedure the author used, however, drastically limits generalizability