Errors of Measurement: Range Restriction

Generalizability is an important principle that cuts across research domains, and it, perhaps, is the best place to begin a discussion of range restriction in communication research. The goal of research in most situations is to draw conclusions about the strength of an effect or size of a relationship among constructs in the population of interest. Identifying findings in the sample utilized in a study is not the end goal, but rather a means to reach the goal of understanding construct tendencies in the population.

There are two obstacles that typically stand in the way in the process of generalizing from a sample to a population: measurement error and range restriction. These two obstacles can be formidable. They almost always result in an underestimate of the real strength of the effect or association in the population, at times by a substantial amount. Failure to take them into account can not only lead to underestimations but even incorrect perceptions about the variables of interest. In the research principles domain, they are referred to as artifacts, a term that implies something that it is not wanted and/or has adverse effects.

To illustrate, say, for instance, that a researcher wants to study communication in first dates, specifically the influence that it has on subsequent interest in a second date. Looking to the literature, the researcher finds the concept of communication competence, which is usually considered via its two main aspects: effectiveness and appropriateness. Effectiveness refers to judgments made by each individual regarding the ability to accomplish his or her goals as a result of the interaction. Appropriateness focuses on assessments of the extent to which the other individual meets expectations and follows social norms. The researcher’s prediction (the alternate hypothesis in research terms) is that the more inappropriate communication the first date partner expresses, the lower the interest in a second date.

Accordingly, this researcher creates and administers two surveys to a sample of college students from his or her institution who recently went on a first date. To assess communication, each is asked to check which, among a series of statements, the other person expressed during the first date. These statements reflect communication that is typically considered inappropriate, including propositions of marriage and inquiries that are highly personal. The total number of checked statements is recorded, which represents the communication variable. To assess the level of interest in a second date, the researcher creates a scale wherein interest is rated from 1 (none whatsoever) to 10 (extremely interested).

Doing the computations, the researcher finds a correlation of −.24 between number of inappropriate statements and level of interest in a second date. As a review of basic statistics, correlations can range from −1.00 to +1.00. The sign (positive or negative) [Page 438]indicates direction, specifically whether larger values on one variable are associated with larger (positive correlation) or smaller (negative correlation) values on the other variable, and vice versa. The first-date correlation is negative as expected because more inappropriate communication should be associated with less interest in a second date. The absolute value indicates how strongly the two variables are associated, and can range from .00 (no association) to 1.00 (a perfect association).

...