Methods for the Detection of Invalid Responses in Survey Data

20/12/2019 13:00
Turkey

The seventh seminar of the Fall Semester of 2019-2020 will be given by Assoc. Prof. Bengü Börkan. The title of her talk is "Methods for the Detection of Invalid Responses in Survey Data."

 

Methods for the Detection of Invalid Responses in Survey Data
Self-administered surveys is widely used in social and behavioral sciences. Research findings consistently indicate that invalid responses in self-administered survey cause misleading results and even wrong conclusions. Nichols, Greene, and Schmolck (1989) define two main types of invalid response. The first response behavior, called content responsive faking, corresponds to responses that is affected by the context and is deliberately misleading. Mischievous responding and socially desirable responding are considered as such a response behavior. The second response behavior, called content nonresponsivity, corresponds to random response patterns. Different terms have been used to refer to this response behavior in the literature. First, it was called random responding (such as Beach, 1989; Berry et al., 1992). Later, researchers realized that invalid nonrandom response patterns were possible as well. Consequently, researchers have adopted more inclusive terms such as inattentive responding (Maniaci & Rogge, 2014; McKibben & Silvia, 2017; Meade & Craig, 2012), careless responding (Curran, Kotrba, & Denison, 2010) and insufficient effort responding (Huang, Curran, Keeney, Poposki, & DeShon, 2012). Participants, who do not cognitively involve in a response process, may display different form of inattentive response behavior due to lack of motivation, absence of opinion or poor questionnaire design (Jin, Chen, & Wang, 2018). Research results shown that as 2.3% or as many as 40% of participants respond inattentively in surveys. Techniques for identifying invalid responses in the data can be categorized as either a priori or post hoc. The prior techniques are planned actions that need to be decided during data collection planning. They often involve adding special items in a survey besides original ones. On the other hand, post hoc techniques could be a planned or unplanned. Unlike prior techniques, they do not required additional survey items to be inserted in surveys; these techniques depends on original survey items to compute indices to detect bias responding. While prior techniques are used to detect both type of inaccurate responding - content responsive faking and content nonresponsivity, post hoc techniques are often used to detect the latter. These commonly used techniques will be discussed in the presentation.