Survey Measurement error 3_Interviewer, respondent, items Flashcards
Ways to estimate interviewer effects
Interpenetrated designs Empirical models (determinants of Rho)
Problems with interpenetrated designs
alter usual field procedures
confound measurement error with non response error
stability of estimates depend on J (# of int)
Logistical issues- attrition, reassignment
Determinants of Rho
- Mode of Interview (Rho lower in self administered mode of admin)
- Computerization (computerization lessen Rho)
- Item format- no diff open/close; factual slightly higher
- Controlled feedback and probing
- Int experience- u shaped curve
- Respondent- older R larger Rho’s
Causal models
Interviewer characteristics
Interviewer behavior
Respondent motivation
Interviewer expectations
Cannell et al.- Tactics
Commitment Feedback Instructions to respondent -Not much evidence - Interviewer as motivator -Proponent of behavior coding
Fowler and Mangione
Impact of training and supervision
Strict standardization
Interviewer variance was a key source of error
Standardized interviewing
goal- remove interviewer as a source of bias
read question as worded
promote comparable data- differences are diff in respondents
Conversational Interviewing
Int and R work together to assure R understands question as intended
-goal- standardize meaning even if word varies