Effect of self-regulated learning strategies on students’ achievement in science: A meta-analysis

Meta-analysis is a quantitative research design that has a variety of purposes that include: a) collecting systematic data from empirical studies (Glass, 1976; Johnson & Christensen, 2008); b) observing the size of the effect that provides the mean difference (Higgerson, 2005; Lipsey & Wilson, 2001); c) analysing estimates of the relationships or construct -level relationships within a population to estimate the parameter values for a population (Hunter & Schmidt, 2004); and d) focusing on the relationship between two different constructs or measurements between treatment groups and control groups (Hedges & Olkin, 1985). In this study, meta-analysis was employed to investigate the impact of self-regulated learning strategies on students' achievement in science subjects.

Literature Research and Data Sources

We used PRISMA (Moher et al., 2009) as a guideline to retrieve the relevant articles. Comprehensive literature searches were performed using the Web of Science, Scopus, JSTOR, ERIC, SpringerLink, Wiley, SAGE, and DOAJ databases that included unpublished and published works. Figure 1 depicts the process of conducting the literature search.

Figure 1. PRISMA flow diagram (Source: Moher et al., 2009)

The following search terms and variants were used in different combinations using the Boolean operators “AND, OR, + and -”, self-regulation, self-regulated, SRL, scientific achievement, self-directed, autonomous, self-regulation strategies, learning strategies, metacognitive strategy, cognitive strategy, motivational strategy, management strategy, biology, chemistry, physics, Science, primary school, higher education, high school, middle school, tertiary level, experiment, quasi-experiment, effect, affect, academic achievement, performance, success, education, educational research, and education.

The search yielded 2,649 results from the Web of Science, 222 from Scopus, 1,336 from JSTOR, 1,817 from ERIC, 464 from SpringerLink, 200 from Wiley, 91 from SAGE, and 10 from DOAJ. However, DOAJ databases did not yield any relevant result because of the lack of research on science subjects compared to mathematics and English. The searches yielded a total of 4,356 studies during the initial screening. For each list, we identified the results based on the title and excluded papers that were unrelated to self-regulated learning and students’ achievement or performance or success. For articles with similar titles, we checked the abstract following the focus of the study. We used references and specific journals such as Metacognitive Journal, Metacognition and Learning Journal, and Cognition to identify other relevant articles.

Inclusion and Exclusion Criteria

We considered three main criteria, namely publication year, publication types and language used based on the suggestion by Kitchenham and Charters (2007) and Okali (2015). First, we selected studies between the year 2000 and 2019. Second, we reviewed only journal articles and third, we accepted papers published in English and Malay. Any item that did not fulfil this criterion was therefore excluded.

Eligibility

We also carried out the second screening process by looking at the title and abstract of the article. If the report was relevant to our study, we read the full article. However, we only chose the item that included statistical synthesis of effect size. At this stage, only 22 studies were selected by using 237 unique data sources.

Coding

Following the initial search and screening, we coded the studies in order to identify the final analytical sample and build the data set. The study characteristics (moderator variables) were coded in order to investigate the possible influence of some of these variables on effect size. A codebook was created specifically for this project using Excel to include the following attributes: a) the name of the author and the year of publication (2000 to 2019); b) subject (science, physics, biology, and chemistry); c) category of SRL, d) SRL strategies, and e) effect size value. Acronym was used to code the SRL strategy and categories of SRL.

Effect Size Extraction and Interpretation

We used two web-based effect size calculators to compute various input of data effect size: (i) Campbell collaboration “http://www.campbellcollaboration.org/escalc/html/ EffectSizeCalculator-SMD-main.php”, and (ii) psychometric “https://www.psychometrica.de/effect_size.html”. Meanwhile, Hattie’s (2009) Continuum of effect sizes (0.2 small, 0.4 medium, 0.6 large) was used to evaluate and explain the effect size.