Summary: | 碩士 === 世新大學 === 行政管理學研究所(含博、碩專班) === 95 === Sample representativeness is essential for telephone polling. However, comprehensive and systematic research sample representativeness and distorting sample in particular, have been rarely seen in Taiwan. With it in mind, this author tries to determine the answer for the causes and effects of distorting samples which were conducted through telephone surveys.
Beginning with literature review, this study identifies four sources of distorting samples. First of all, ambiguous definition of population and election of sampling frame can contribute to insufficient and inappropriate coverage rate. In addition, it is reasonable to claim there is difference with or without sampling within household whenever telephone survey is conducted. As far as polling process is concerned, visiting date, visiting period and skill of interviewer have been arguably made difference. Last but not the least, both the sensitiveness of the subject, and the length of the questionnaire can cause non-response problems.
Sampling sources for this research come from two domestic telephone survey institutions concerning 2004 presidential election, in lieu of random digit dialing and sampling within household. An extensive examination of 74 telephone survey data banks highlights unexpectedly acute distortions between telephone survey sample and population.
By using various weighting methods against election results, this author argues that weighting has been a common tool getting desirable sample representativeness in telephone surveys, but the effect of weighting is finite. It therefore suggests that it could be better to callback and model weights other than weighting for coming up representative samples.
|