Optimizing measurement in Internet-based research : Response scales and sensor data

Abstract

The present PhD thesis lies within the area of Internet-based research; specifically it is concerned with Internet-based assessment via questionnaires and smartphones. The thesis investigates the influence of response scales and objective sensor data from smartphones on the data gathering process and data quality. Internet-based questionnaires and tests are becoming increasingly common in psychology and other social sciences (Krantz & Reips, 2017; Wolfe, 2017). It is therefore important, for researchers and practitioners alike, to base decisions about their research design and data gathering process, on solid empirical advice. The first research article compared two types of response scales, visual analogue scales (VASs) and Likert-type scales, with regard to non-response. Results indicated a positive effect of VASs with regard to lowering dropout of participants. The second research article investigated the validity of measurement, again comparing VASs and Likert-type scales. Results largely indicated measurement equivalence between the two response scale versions, with some evidence for better measurement quality with VASs. The third research article investigated the validity of objective sensor data in an experience sampling design. The association of subjective well-being with smartphone tilt was investigated in two separate samples. Results provided evidence for the validity of smartphone tilt as an indicator of subjective wellbeing. Furthermore, potential biases and problems when implementing objective data are discussed, specifically when different software implementations and operating systems are involved. In conclusion, the PhD thesis offers valuable insights on Internet-based assessment. The VAS’s position as a superior response scale was strengthened. Its advantages over more traditional Likert-type scales, e.g., offering better distributional properties and more valid information, were confirmed and no disadvantages emerged. Smartphone sensor data were shown to provide a way to validate self-report measurement, if potentially important caveats related to differences in data are identified and addressed.