Publication Date

2013-05-08

Availability

Embargoed

Embargo Period

2015-05-08

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PHD)

Department

Communication Studies (Communication)

Date of Defense

2013-04-08

First Committee Member

Michael J. Beatty

Second Committee Member

Diane M. Millette

Third Committee Member

Don W. Stacks

Fourth Committee Member

George Wilson

Abstract

It has long been known that measures can vary in the degree to which they are congeneric. The criteria for congenericity are important because (1) estimates of scale reliability such as those generated by Cronbach’s alpha greatly overestimate the reliability of noncongeneric measures, and (2) error covariances indicate that latent variable(s) other than the construct of interest contribute to item scores, thereby producing confounded measurements. Despite the observation that most measures in psychological research, which are similar to many communication measures, are noncongeneric, measures in communication research have been treated as though they are congeneric. In the present study, four measures frequently used in communication research—McCroskey’s (1982) Personal Report of Communication Apprehension (PRCA-24), Wiemann’s (1977) Communicative Competence Scale (CCS), Infante and Rancer’s (1982) Argumentativeness Scale (ARG), and Infante and Wigley’s (1986) Verbal Aggressiveness Scale (VAS)—were submitted to confirmatory factor analysis and examined for congenericity. When basic congenericity was met, the measure(s) were also tested for tau-equivalence and parallelism, both of which place more restrictive requirements on measurement models. In cases where noncongenericity was indicated, recommendations were made for the treatment of noncongeneric measures in ways that account for nonrandom error and provide a more accurate picture of the measure’s actual reliability and validity as indicated by the data.

Keywords

Interpersonal communication measures; confirmatory factor analysis; classical test theory; congenericity and measurement error; correlated uniqueness models; reliability

Share

COinS