Research Information System
Statistics in Transition New Series, vol.22, no.2, pp.67-93, 2021 (Scopus)
In surveys, non-sampling errors, due to their complex nature, are more challenging to quantify compared to sampling errors. Avoiding the release of these errors, however, results in biased survey estimates. In our previous paper, we devised the best interviewer allocation technique by using a nested experimental design to study response error estimation. In this study, in order to illustrate the effectiveness of this methodology in a different context, we apply it in interview-reinterview surveys relating to the time use and life satisfaction of academicians at Middle East Technical University, Turkey. An analysis of the pilot survey data showed that only half of the data was reliable, while the other half revealed interviewer effects. Prior to the main survey, interviewers underwent training in the course of which particular emphasis was put on the above-mentioned questions. In effect, the previously observed response variances which accounted for the total variance and data unreliability, were reduced considerably, increasing the quality of the main survey.