Desktop version

Home arrow Philosophy

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Current and future directions for variable-centered approaches to strategic processing research

In sum, variable-centered analyses are statistical methods used to prepare or analyze data from previously identified groups (Laursen & Hoff, 2006). Whereas the goal of using person-centered analyses (e.g., cluster analysis and latent class analysis) is to identify previously unobserved groups based upon common patterns within the data, the goal of variable-centered analyses is to explore, compare, or contrast extant groups within the data. Data aggregation prepares data for analyses. The analyses we outlined in this chapter include GLM analyses (i.e., t-tests, ANOVA, multiple linear regression), count models, path analysis, structural equation modeling, and growth curve models.

In this chapter, we explained the purpose of and process for using variable-centered analyses to study strategy use and strategic processing. Researchers were able to answer key research questions about strategy use through variable-centered analytic methods. For example, Bernacki and colleagues’ (2012) study used path analysis to analyze trace data from college students’ use of a technology-enhanced environment to study relationships among strategy use, self-reported motivation, and learning performance. Using path model analysis meant they were able to discover that approach-based motivation was positively related to strategy use, and avoidance-based motivation was negatively related. Additionally, variable-centered analyses provided other affordances for research, such as analyzing use of count data. Sometimes studies of strategy use involve counts of the number of times participants use particular strategies. These can be counts of strategy use via observation (e.g., Hagen et al., 2014), participants’ think-aloud verbalizations of strategy use (Greene et al., 2018), or from trace data from computer-based learning environments (Bernacki, 2018). Researchers can analyze count data that is non-normally distributed through generalized linear model analyses. Also, variable-centered analyses can be used to model complex relations among numerous measured variables (e.g., path analysis), to understand latent variables and structures (e.g., SEM), as well as how strategic processing changes over time (e.g., repeated-measures ANOVA, growth modeling). Variable-centered analyses of strategic processing data have been and likely will continue to be a prominent way of understanding how people enact strategies, as well as the precursors and consequences of such enactment.

Future Directions for Research

There are opportunities to expand the use of variable-centered analyses for strategy use in future research. We identified many researchers who utilized self-report surveys and questionnaires about strategy use (e.g., Askell-Williams et al., 2012; Mirzaei et al., 2014; Ruffing et al., 2015). However, strategy use is dynamic and not stable. Researchers studying strategy use that incorporates behavioral data have an opportunity to capture learning behaviors as they occur, via think-aloud protocols (Anmarkrud et al., 2013), eye fixations on a page (Arya & Feathers, 2012), or other methods. Capturing these strategic processing behaviors as they occur can be more useful than self-report data because it allows for modeling the dynamic nature of strategy use, and can afford an understanding of the sequential and contingent nature of strategic processing (Ben-Eliyahu & Bernacki, 2015; Binbasaran Tiiysuzoglu & Greene, 2015). In the future, variable-centered analyses, such as path modeling, could be used to understand how people leverage feedback to dynamically adjust the depth of their strategy use (Dinsmore, 2017), such as when frequent difficulty with formative assessments leads to participants shifting from surface- to deep-strategy use.

Often, researchers have studied how strategy use predicts achievement or learning outcomes (e.g., Carr et al., 2011). However, there are fewer studies that exist where researchers have considered strategy use itself as the outcome variable (e.g., Greene et al., 2011; Vasilyeva et al., 2015). In the future, researchers may study strategy use as an outcome in itself more often, rather than establishing the relationships between strategy use and other variables (e.g., GPA). Given the established relationship between strategy use and performance outcomes, more research is needed on how to foster effective strategic knowledge and processing among those who would otherwise struggle to enact without support (Bjork, Dunlosky, & Kornell, 2013).

In the future, we encourage diversifying methodological and analytical techniques in the field of strategy use. We identified many studies that utilized quasi-experimental designs (Aghaie & Zhang, 2012) or experimental studies (Cantrell et al., 2014), but fewer case studies. It may be that qualitative research is needed to better understand why people do and do not enact effective strategic processing, and what variables are most important to capture and analyze (e.g., self-efficacy for strategy use; Zimmerman, 2013). Likewise, mixed methods might enhance the knowledge base and research quality in the field of strategy use. For example, researchers might use mixed methods to iterate between what participants actually do when studying (i.e., quantitative data derived from observation) and what they think they are doing (i.e., qualitative investigations of participants’ impressions and experience). Such mixed methods research designs may afford greater insight into the variables associated with proper calibration and metacognitive knowledge during strategic processing (Pieschl, Stallmann, & Bromme, 2014).

In addition to methodological and analytical diversification, diversification of samples is also an area of further exploration. For example, Vasilyeva et al. (2015) analyzed cross-cultural differences in strategy use between US and Taiwanese students. They found that Taiwanese children used retrieval practice more than US children. Such research is necessary to understand how strategy use varies across contexts, and why. However, such work must necessarily be conducted from a perspective derived from within the culture, rather than imposing ideas derived from one culture onto another (King & McInerney, 2018).

Conclusion

In conclusion, strategy use data comes in different forms, and some of these data can be analyzed with variable-centered analyses. The variable-centered analyses described here include general linear model approaches such as correlational and regression analyses, path analyses, structural equational models, and growth models. Data aggregation can be used to make strategy use data more amenable to variablecentered analyses, although in some cases count-based analyses must be used. Though the variable-centered analyses most appropriate for a given situation are determined by the research question, the type of data collected, and the goal of the research methods used, we demonstrated in this chapter how variable-centered analyses can be used to answer questions about the nature of strategy use and its relationships with other variables. Specifically, we look forward to seeing how researchers incorporate new applications of analytic methods for understanding strategic processing data.

 
<<   CONTENTS   >>

Related topics