fredag 9 november 2018

Evaluating and Reducing Biases in Mixed Mode Survey Data, November 15, 2018

The Next AAPOR Webinar
In recent years survey researchers have shown increased interest in using mixed-mode designs that combine more than one mode of administration for data collection in a survey. The primary motivations for using mixed-mode include increasing response rates, reducing selection bias, and saving on survey administration costs. A well-known problem of mixed-mode designs is that survey questions are associated with different measurement error bias when posed under different modes. These differences are a risk for comparability between population subgroups and in time. Survey practitioners using mixed mode survey data, therefore, are interested in understanding how estimates obtained from mode-specific respondent sets are affected by mode differences in selection or measurement bias. In this course, we describe statistical methodology that can be used to estimate and to reduce these bias components. We focus on the data collection design stage and the estimation stage, and assume questionnaires as optimized and fixed. We start by the definitions of selection and measurement effects and describe how these relate to survey biases. Subsequently, we give overviews on covariate based estimation, the instrumental variable method, the re-interview method, and time-series stabilization. Particular attention is given to the assumptions of the estimation approaches to enable practitioners to decide which methodology is suitable for their design.

Registation and read more....