Mode-specific measurement biases and selection biases in health statistics
The discussion paper provides estimates of mode-specific measurement biases for key health statistics between web and interviewer-assisted data collection. Biases for some statistics are sizeable and imply that analysis of mixed-mode health statistics should be performed with care.
Mixed-mode adaptive survey designs have become a standard in many survey settings. Also the combination of mixing modes and adapting effort has been elaborated and applied. However, the focus has been on balancing cost and nonresponse. In this paper, we show that mode-specific measurement biases can be large and cannot be ignored.
