Meta Analysis

Meta-analysis is the process of systematically, objectively, and statistically combining results from independent quantitative studies on a specific topic to reinterpret them. Known in the literature as “the analysis of analyses,” this method aims to achieve more general and reliable conclusions by aggregating quantitative data from studies sharing common characteristics, rather than examining individual research findings in isolation.

History and Development

The use of analytical procedures to combine research findings dates back to the early 20th century, particularly to Pearson’s (1904) work aggregating correlation coefficients. However, the term “meta-analysis” was first introduced by Gene V. Glass in 1976. Glass emphasized that the defining feature distinguishing this method from classical literature reviews is its foundation in statistical analysis. Subsequently, scientists such as Hedges & Olkin (1985), Hunter & Schmidt (1990), and Rosenthal (1991) strengthened the theoretical and practical foundations of meta-analysis techniques.

Need for Meta-Analysis

Today, the number of scientific studies is increasing rapidly, and these studies often produce conflicting results. This situation makes it difficult for decision-makers and policymakers, especially in applied fields, to access reliable information. Classical literature reviews are typically descriptive in nature, susceptible to the researcher’s subjective interpretation, and have low sensitivity to publication bias. In contrast, meta-analysis overcomes these limitations through powerful statistical tools and offers a more objective means of summarizing scientific knowledge.

Key Concepts and Stages

The meta-analysis process consists of seven key stages:

1. Definition of the problem
2. Determination of inclusion criteria
3. Collection of relevant studies
4. Coding and classification of studies
5. Calculation of effect size
6. Analysis of data and model selection (fixed vs. random effects model)
7. Interpretation and reporting of results

Effect Size

In meta-analyses, the unit that enables comparison across studies is the “effect size.” Cohen’s d (1988) is one of the most commonly used measures in this context, calculated by dividing the difference between the means of the experimental and control groups by the standard deviation. Effect size provides insight into the direction and magnitude of an intervention’s impact.

Fixed and Random Effects Models

Two primary statistical models are used in meta-analysis:

• Fixed effects model assumes that all studies share the same true effect size.
• Random effects model acknowledges that effect sizes may vary across studies. This model is preferred when making broader generalizations.

Publication Bias and Corrective Methods

Meta-analyses conducted solely on published studies are vulnerable to publication bias, as studies with statistically significant results are more likely to be published. This phenomenon is also known as the “file drawer problem.” Various methods such as Rosenthal’s “failsafe N,” Orwin’s adjusted calculations, Egger’s test, and funnel plots can be used to measure and mitigate this bias.

Data Collection, Coding, and Analysis Tools

In meta-analysis studies, selected studies are coded using software such as SPSS and Excel according to specific criteria. During this coding process, independent variables (sample size, duration of intervention, measurement instrument, etc.) and dependent variables (achievement, attitude, anxiety, etc.) are taken into account. Each study is analyzed based on its effect size and variance, and the structure of the dataset is evaluated using homogeneity tests (e.g., Q test).

Application Areas and Contributions

The meta-analysis method is widely used in fields such as medicine, education, psychology, and economics. It is particularly favored in developing educational policies, comparing teaching methods, and conducting program evaluations. Meta-analysis not only synthesizes past research but can also guide the design of new studies. If previous meta-analyses have shown that certain variables have no significant effect, excluding these variables from future research can save time and resources.

Meta-analysis has become an indispensable method in modern research for systematically, objectively, and scientifically integrating knowledge. It is a vital tool in research synthesis, enhancing the generalizability of findings and clarifying contradictory results in the literature. However, this method must be applied carefully and correctly, with rigorous selection of data and interpretations grounded in context.