PGR Student Presentation - Using paradata to explain the interviewer contribution to survey design effect Seminar
- Time:
- 13:00
- Date:
- 13 May 2011
- Venue:
- Room 7033 (7C), Building 54 Highfield Campus
For more information regarding this seminar, please telephone Dr Claire Bailey on +44 (0)23 8059 2577 or email C.E.Bailey@soton.ac.uk .
Event details
Social Statistics and Demography Seminar Series
Using paradata to explain the interviewer contribution to survey design effect The interviewer is a well recognized source of survey error. In this paper I focus on the interviewer contribution to measurement error and, more precisely, on the variability of survey estimates that is introduced by the interviewer: the interviewer effect. Face-to-face interview surveys generally employ a clustered sample design, in which geographical clusters are first selected and then individuals or households are selected within clusters . This design can lead to inflation of the variance of survey estimates, relative to a simple random sample, due to the greater similarity between respondents in the same cluster than is evident in the population as a whole. This phenomenon is referred to as the design effect. Because, usually, there is only one interviewer working in each geographical cluster it is difficult to separate the design effect due to areas from that which is caused by interviewers.
I use a cross-classified multilevel model to disentangle area and interviewer contributions to the design effect for a range of variables in the UK National Travel Survey. Additionally, I look at various interviewers? characteristics to asses to what extent they can explain the variability of the survey estimates when controlled for the area (cluster) and respondent?s characteristics. In other words I seek to understand how variance in a range of survey outcomes is related to the characteristics of individual interviewers.
Speaker information
Gosia Turner ,A consultant employed by the Higher Education Academy.