Desktop version

Home arrow Philosophy

  • Increase font
  • Decrease font

<<   CONTENTS   >>

Variable-centered analysis techniques

Data from a variety of quantitative research designs (e.g., true experimental, quasi-experimental, non-experimental) can be analyzed using variable-centered techniques. Variable-centered analyses can be conducted in studies of strategy use across different time frames as well, including during one learning episode (e.g., Greene et al., 2018) or over multiple episodes (e.g., Carr et al., 2011). Further, variable-centered analyses have been used to understand strategic processing across a variety of contexts, from labs to classrooms to learning online. Strategy use can either be studied as an independent or a dependent variable within variable-centered processes. Before analyses begin, however, it is often necessary to aggregate strategy use data (Greene, Dellinger, Binbasaran Tüysüzoglu, & Costa, 2013).

Data Aggregation

Researchers who study strategy use and strategic processing sometimes need to utilize data aggregation before analyzing their data. There can be many strategies observed in a sample, often more than ten (e.g., various memorization strategies, higher-order strategies; Dunlosky et al., 2013). In such cases, it can be a challenge to use variablecentered analyses to compare the efficacy of those strategies to one another, due to sample size needs. For example, if 13 strategies are observed in a sample, variable-centered analysis (e.g., research question: which of the 13 strategies is the strongest predictor of learning performance?) guidelines would suggest the need for a relatively large sample (i.e., 117 per Green, 1991). Aggregation can be used to address such challenges, particularly when the use of specific strategies is less important than whether particular types of strategies were used at all. For example, it may not matter whether one participant elaborates, another spaces practice, or a third participant self-tests. What matters is the number of times they invoke any of these effective strategies (Dunlosky et al., 2013). In this situation, it may be useful to create a macro-level aggregate variable (e.g., deep strategy use) comprised of the sum of the frequency of use of these micro-level strategies. In one study, micro-level strategy use data (e.g., frequency of summarizing) were aggregated into surface- and deep-strategy use macro-level variables in order to predict differences in learning outcomes (Deekens et al., 2018). In this study it was less important which specific strategy was used (e.g., taking notes, summarizing, etc.) than the number of times a participant invoked each type of strategy. In another study, the authors aggregated micro-level strategy use variables (e.g., type of note taking strategy) in order to describe the variety of strategies used in their given case or group (Hagen, Braasch, & Braten, 2014). They found that intertextual knowledge elaboration use statistically significantly predicted deep-level comprehension outcomes when reading to construct an argument, whereas this relationship was not present in participants who used this strategy while reading to summarize.

Aggregation can be used to test posited relations in models of strategy use or selfregulated learning (SRL; Greene & Azevedo, 2009). Many of these models are conceptualized at the macro-level. For example, in Zimmermans model of SRL, at a macro-level planning in the forethought phase is posited to drive strategy use in the performance phase (Zimmerman, 2013). Yet the data collected are often at the micro-level (e.g., a participant makes a subgoal as one kind of planning, whereas another student calibrates a task definition; one student uses an elaboration strategy during performance, whereas another uses highlighting), thus aggregating to the macro-level is necessary to test whether indeed macro-level planning predicts macro-level strategy use (Greene et al., 2013). In this way, micro-level data can be aggregated into macro-level data to test hypothesized relations in the model (e.g., changes in task understanding can affect the strategies used).

Data can also be aggregated by time or learning phase. For example, data from think-aloud protocols can be aggregated into types of strategies used during learning, as opposed to before or after learning (Greene, Robertson, & Costa, 2011). In sum, data aggregation allows for different, additional, or further analysis of data that can help inform research on strategy use. Once data have been aggregated, they can be analyzed using variable-centered techniques, just as non-aggregated data can be analyzed. In the remainder of this section, we describe various variable-centered analysis techniques, with examples of each to illustrate how they differ.

<<   CONTENTS   >>

Related topics