H K Suen | PubMed Health

PubMed1993

A controlled randomized response technique.

The randomized response survey technique appears to be suitable for studies of sensitive sexual behaviors, particularly in AIDS-related research. However, existing methods provide only estimates of group statistics, not of individual information. Additionally, the popular "unrelated question" approach requires the knowledge of the parameters of the unrelated question. In this article, a variation of the unrelated-question method is suggested for use. Specifically, it is suggested that the unrelated question be one to which the response is known to be "yes." Through this "controlled" approach, the raw data become a direct linear transformation of the response to the sensitive question, and thus can be used directly in regression and other analyses at the individual score level. The estimation of the parameters for the unrelated question is not necessary and the hesitation to provide a "yes" response found in the "forced choice" method is minimized.

Female↗

PubMed1984

A Bayesian aggregate meta-analytic evaluation approach.

The Bayesian inferential process is modified to be used as an approach to an aggregate meta-analytic evaluation. Comparisons of the Bayesian approach with the traditional average effect size meta-analytic approach indicated that the Bayesian approach was more sensitive to differences between studies. The approaches to derive descriptive and inferential statistics were more consistent in the Bayesian approach. Because of its ability to account for all available information, the Bayesian approach was statistically more powerful. It is recommended that this approach be used for combining evaluation results when primary data are not available and when all evaluations involve comparisons of two independent samples.

Evaluation Studies as Topic↗