The use of process data to examine reading strategies

UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
Juanita C. Hicks (Creator)
The University of North Carolina at Greensboro (UNCG )
Web Site:
John Willse

Abstract: Researchers are increasingly interested in the cognitive behaviors students display during tests. This interest has led researchers to look for innovative ways to collect this type of data. Due to the proliferation of computer-based assessments, process data has become popular for its ability to help show what students know, what students don’t know, and how students interact during assessments. Aim: The aims of the current study are 1) to use process data to identify potential reading strategies and 2) to examine if reading strategy is associated with gender, race/ethnicity, and differences in performance. Methods: Apply latent profile analysis (LPA) to extracted process data variables collected from US examinees who participated in the literacy section of the Program for the International Assessment of Adult Competencies (PIAAC). The variables are item response time and number of highlight events per item. Results: A two-class solution provided the best fit for the data in each testlet of the literacy section of the PIAAC. Class one progressed through items in each testlet faster than class two. Class one most closely resembled a skimming strategy while class two most closely resembled a full-reading strategy. However, there was not conclusive evidence to suggest that the classes were reminiscent of skimming and full-reading. Class assignment had no significant relationship with gender nor race/ethnicity, and there was no significant difference in literacy performance between the two classes, except in one case. Even then, both classes performed at a level two on the PIAAC literacy achievement scale. Discussion: Response time was found to be the only discriminating variable in the identification of patterns related to reading strategies. While there was some separation between classes, it was minimal in some cases. Response time was found to be useful but not enough to identify conclusive reading strategies. Further research is needed to identify process data variables with explanatory power other than response time to aid in the identification of reading strategies.

Additional Information

Language: English
Date: 2019
Latent Profile Analysis, PIAAC, Process Data, Reading Strategies
Test-taking skills
Learning strategies
Educational tests and measurements

Email this document to