Simulating Response Latitude Effects in Attitude Surveys using IRT
Date of Award
Doctor of Philosophy (Ph.D.)
Scott Highhouse (Committee Member)
John Tisak (Committee Member)
James Albert (Committee Member)
A response latitude represents the range of response options that a respondent is willing to endorse in response to a scale item. For any particular attitude survey, some respondents should naturally feel opinionated and involved with the topic, and therefore expend the mental energy needed to provide careful, precise responses. This results in a narrow response latitude and relatively favorable scale psychometric properties. Other respondents, however, probably lack a discernible opinion about the topic, and provide imprecise or careless responses. This results in wide response latitudes and relatively unfavorable psychometric scale properties. The present study is a Monte Carlo simulation, wherein response latitude width, the proportion of wide response latitudes in a sample, and sample size were systematically varied to understand the role each plays in terms of psychometric scale quality. Results indicate that response latitude width had a much greater impact on scale quality than did the proportion of wide response latitudes in a sample and sample size. Response latitudes equal to or greater than 1.50 resulted in poor psychometric quality across many quality indicators such as model-data fit and test information. Finally, internal consistency, internal validity, and external validity estimates were all relatively robust to wide response latitudes and suggest that latitude width may have a somewhat limited effect on the measurement of psychological constructs.
Lake, Christopher, "Simulating Response Latitude Effects in Attitude Surveys using IRT" (2014). Psychology Ph.D. Dissertations. 138.