Variable-Centered Approaches to Analyzing Strategic Processing Data

• In Event: Investigating Strategies and Strategy Use: Where Do We Go From Here?

Abstract

Strategy use, or strategic processing, is dynamic and malleable, as well as associated with achievement in higher education (Dinsmore, 2017; Schneider & Preckel, 2017). Researchers have worked to identify strategies that promote quality learning and achievement outcomes (e.g., practice testing and distributed practice) and then encourage students to use them. Due to the dynamic nature of strategic processing, within and between domains and tasks, research methods that are capable of capturing dynamic processing are necessary (Dinsmore & Zoellner, 2018). However, such research methods can produce a lot of data, which can be challenging to analyze. These data may be analyzed in many ways, including variable-centered analyses (e.g., correlation, regression, path analysis, structural equation modeling, and growth models; Laursen & Hoff, 2006).
Variable-centered statistical methods involve analyses of relations between variables to produce a summarization of these relations, within a given set of parameters, to describe the entire population (Howard & Hoffman, 2018). Data from a variety of quantitative research designs can be analyzed using variable-centered techniques (e.g., true experimental, quasi-experimental, non-experimental). Variable-centered analyses have been used to understand strategic processing across a variety of contexts, from labs to classrooms to learning online. Variable-centered analyses have also been used to understand strategic processing across a variety of time frames, such as during one learning episode or over multiple episodes. Additionally, strategy use can either be studied as an independent or a dependent variable within variable-centered processes.
One particular challenge associated with using variable-centered analyses with strategy use data is that learners can use many different strategies, therefore an analysis of which strategies (e.g., frequency of note-taking, frequency of elaboration) are most effective can require unreasonably large sample sizes (Green, 1991). Thus, before analyses begin, it is often necessary to aggregate strategy use data (Greene, Dellinger, Binbasaran Tüysüzoğlu, & Costa, 2013). Aggregation can be used to reduce the number of strategy variables analyzed, particularly in instances when whether or not participants use specific strategies is less important than whether or not participants used effective or ineffective strategies overall.
Another challenge using variable-centered analyses with strategic processing data is that many researchers investigate whether counts of strategy use relate to other variables (e.g., demographics, treatments, learning outcomes). Often, such count data are not normally distributed (Gall, Gall & Borg, 2007; Greene et al., 2011). Thus, if non-normally distributed strategy use count data are used as a criterion variable, a basic assumption of general linear model analyses is violated. In such cases, researchers must use statistical techniques that can compensate for these non-normal distributions of data (i.e., Generalized Linear Model analyses).
In this work, we will describe the purposes and relations among various variable-centered analysis techniques, as well as when they should be used. Also, we discuss how data aggregation and alternative statistical techniques can be used to address the concerns outlined above. We discuss future areas of growth for variable-centered approaches in the area of strategic processing, and we encourage diversifying methodological and analytical techniques used to understand strategic processing.

Authors

• Rebekah Freed, University of North Carolina - Chapel Hill

• Jeff A. Greene, University of North Carolina - Chapel Hill

• Robert D Plumley, University of North Carolina - Chapel Hill