Date Created

5-2014

Document Type

Dissertation

Department

College of Education and Behavioral Sciences, Applied Statistics and Research Methods, ASRM Student Work

First Advisor

Schaffer, Jay Ryan, 1969-

Abstract

Estimating prevalent rates of sensitive behaviors using self-report measures generally resulted in bias estimates when direct questioning approaches were used. Random Response (RR) and Non-random Response (NRR) models were developed to provide an additional layer of confidentiality that was meant to illicit more honest reporting. Despite these efforts, there was evidence that survey participants using these techniques do not always report honestly, and as a result, estimates from these techniques were biased. The current study examined the statistical efficiency, using the ratio of MSE between the RR models, the unrelated question technique (UQT) and forced choice technique (FCT) and the NRR models, the item count technique (ICT), double item count technique (DICT) and the single sample count technique (SSC). Simulations of a large range of sensitive prevalent rates and sample sizes were performed where estimates were compared in terms of increasing levels of non-compliance. In addition, for NRR models exclusively, techniques were compared similarly by list size (3-item, 4-item, and 5-item) as well as between each of the NRR models (ICT, DICT, and SSC). Results of the study indicated that the UQT optimal model was the most efficient of the techniques in the presence of equivalent non-compliance rates. However, if the DICT optimal 5-item model improved compliance, this model became more efficient depending on the sensitive prevalent rate estimated and the sample size. As a result, the study demonstrated that in certain situations, the non-random response double item count technique optimal model was as or more efficient than the random response unrelated questioning technique optimal model. Applications of the findings and the development of general guidelines were discussed.

Abstract Format

html

Keywords

Random data (Statistics); Linear models (Statistics); Social surveys; Random response techniques; Indirect question techniques

Extent

278 pages

Rights Statement

Copyright is held by author.

Digital Origin

Born digital

Share

COinS