LEARNING ANALYTICS AND ITS DATA SOURCES: WHY WE NEED TO FOSTER ALL OF THEM

Research output: Chapter in Book/Report/Conference proceedingConference article in proceedingAcademicpeer-review

Abstract

The search for rigor in learning analytics applications has placed survey data in the suspect’s corner, favoring more objective trace data. A potential lack of objectivity in survey data is the existence of response styles, the tendency of respondents to answer survey items in a particular biased manner, such as yeah saying or always disagreeing. Making use of multiple survey instruments that exhibit similar types of response styles, our empirical study identifies response style bias by estimating the aggregate level of a set of response styles, amongst them the Acquiescence Response Style and the Dis-Acquiescence Response Style. We next demonstrate that trace variables are indeed bias-free in that their estimated response style components are small in size, accounting for minimal explained variation. Remarkably, course performance data is not bias-free, implying that predictive modelling for learning analytics purposes will, in general, profit from the inclusion of these bias components or apply survey data containing such response style bias to increase predictive power.
Original languageEnglish
Title of host publicationProceedings of the IADIS International Conference Cognition Exploratory Learning in the Digital Age
EditorsDemetrios G. Sampson, Dirk Ifenthaler, Pedro Isaías
Place of PublicationLisbon
PublisherIADIS Press
Pages123-130
Number of pages8
Volume18
ISBN (Print)978-989-8704-33-7
Publication statusPublished - 2021

Publication series

SeriesCELDA Proceedings

Keywords

  • Learning analytics
  • trace data
  • survey data
  • course performance data
  • response styles

Cite this