Creswell (2008) defines a meta-analysis as "a type of research report in which the author integrates the findings of many (primary source) research studies" (p. 642). Completing a meta-analysis involves establishing research criteria, an exhaustive search of the literature, coding each appropriate article for researcher-determined variables, and compiling the findings into one concise document with its own conclusions.
Graham, S., & Perin, D. (2007). A meta-analysis of writing instruction for adolescent students. Journal of Educational Psychology, 99, 445-476.
Abstract [with notes added in red]:
There is considerable concern that the majority of adolescents do not develop the competence in writing they need to be successful in school, the workplace, or their personal lives. A common explanation for why youngsters do not write well is that schools do not do a good job of teaching this complex skill. In an effort to identify effective instructional practices for teaching writing to adolescents, the authors conducted a meta-analysis of the writing intervention literature (Grades 4-12), focusing their efforts on experimental and quasi-experimental studies [The researchers did an extensive search to find all articles that met these criteria]. They located 123 documents that yielded 154 effect sizes for quality of writing [The researchers found 123 documents that met their criteria. Meta-analyses could include a smaller number of studies, depending on the research questions, criteria, and the breadth of the given area]. The authors calculated an average weighted effect size (presented in parentheses) for the following 11 interventions: strategy instruction (0.82), summarization (0.82), peer assistance (0.75), setting product goals (0.70), word processing (0.55), sentence combining (0.50), inquiry (0.32), prewriting activities (0.32), process writing approach (0.32), study of models (0.25), grammar instruction (- 0.32) [Meta-analysts use effect sizes as a statistical way to identify the strength of the conclusions and/or the strength of the variables].