ABSTRACT
This chapter presents a number of options to include measurement error, which is the result of one or more deficiencies in the answering process, either consciously by lack of motivation, unconsciously by insufficient cognitive ability, or both. Linear optimization problems can be solved analytically and do not require numerical approximation. Second best is that functions are nonlinear but convex. Convexity still is tractable and various numerical methods exist to find global optimal points. A response quality indicator, also termed data quality indicator, summarizes respondent answering behavior that is a sign of an increased risk of measurement error for at least a subset of the survey variables. It is possible but not at all customary to contact respondents when a risk of measurement error is detected. However, in a panel or cohort setting response quality in previous waves may be used to inform decisions. ASDs focuses directly on the combined effect of nonresponse and measurement error against a specified benchmark design.
