Differences in Univariate Values Versus Multivariate Relationships.

Abstract

A recent debate has called to light an important issue for communication researchers and other social scientists: Does the use of student samples jeopardize the validity of research? Importantly, this question has failed to make a critical distinction between estimates of univariate values versus those examining multivariate relationships. Estimates of univariate values provide information on the prevalence of an attitude or belief from the sampled population to the overall population. Estimates of multivariate relationships, however, provide information on the validity of our theories. This distinction is examined in a study of identification with Diana, Princess of Wales. Results show that although the estimates of univariate values differed across three samples, the multivariate relationships between the variables were more stable, and therefore the underlying process appears to hold. We argue that this finding illustrates the distinction between estimates of univariate means versus multivariate relationships and is evidence that nonprobability samples can be important for many reasons, including testing the consistency of processes.

Lies in the Eye of the Beholder: Asymmetric Beliefs about One’s Own and Others’ Deceptiveness in Mediated and Face-to-Face Communication

“…A sample of college students was selected to test the hypotheses. Although college student samples are problematic in making descriptive inferences about a population, they are useful in investigating relationships between theoretically derived variables (Basil, Brown, & Bocarnea, 2002; Li, 2008; Mook, 1983), such as self-other comparisons and media—face-to-face comparisons. College student samples are widespread in testing the self-other discrepancy theoretical framework (e.g., Alicke & Govorun, 2005; Kruger & Dunning, 1999; Pronin et al, 2002).…”

An Examination of Consumer–Company Identification as a Key Predictor of Consumer Responses in Corporate Crisis

“…For example, changes in attitudinal variables may be better captured through a longitudinal study. Second, although using a student sample can be useful in testing theoretically linked relationships (Basil, Brown, & Bocarnea, ), the use of a student sample in this study may bring some limitations regarding generalizability. Future research could employ consumer panels to better represent consumer population.…”

Third-Person Effect, Optimistic Bias, and Sufficiency Resource in Internet Use

“…Therefore, a random sample of general population is not crucial in testing multivariate relationships; (b) A student sample could not be used to produce generalizable results to the whole population but could reveal multivariate relationships as well as a random sample from the general population. The results about multivariate relationships from a student sample were found consistent with those from a random sample of the general population (Basil, Brown, & Bocarnea, 2002); (c) If a theory is true in the population, it should also apply to and stand the test through a student sample. With a carefully designed questionnaire, it is expected that the study produces meaningful data and offers insights to understanding the relationships between sufficiency resource and Internet use, the third‐person effect, and the optimistic bias in Internet communication.…”