In lieu of an abstract, here is a brief excerpt of the content:

GLOSSARY A Posteriori Tests: Tests that are conducted after examination of study results. Used to explore patterns that seem to be emerging in the data. The same as Post Hoc tests. A Priori Tests: Tests that are planned before the examination of the results of the studies under analysis. The same as Planned Tests. Aggregate Analysis: The integration of evidence across studies when a description of the quantitative frequency or level of an event is the focus of a research synthesis. Agreement Rate: The most widely used index of interrater reliability in research synthesis. Defined as the number of observations agreed on divided by the total number of observations. Also called percentage agreement. Apples and Oranges Problem: A metaphor for studies that appear related, but are actually measuring different things. A label sometimes given as a criticism of meta-analysis because meta-analysis combines studies that may have differing methods and operational definitions of the variables involved in the calculation of effect sizes. Artifact Correction: The modification of an estimate of effect size (usually by an artifact multiplier) to correct for the effects of an artifact. See also Artifact Multiplier. Artifact Distribution: A distribution of values for a particular artifact multiplier derived from a particular research literature. Artifact Multiplier: The factor by which a statistical or measurement artifact changes the expected observed value of a statistic. Artifacts: Statistical and measurement imperfections that cause observed statistics to depart from the population (parameter ) values the researcher intends to estimate. ArtificialDichotomization: Thearbitrarydivisionofscores on a measure of a continuous variable into two categories. Attenuation: The reduction or downward bias in the observed magnitude of an effect size produced by methodological limitations in a study such as measurement error or range restriction. Available Case Analysis: Also called pairwise deletion, a method for missing data analysis that uses all available data to estimate parameters in a distribution so that, for example, all available pairs of values for two variables are used to estimate a correlation. Bayes Posterior Coverage: A Bayesian analogue to the confidence interval. Bayes’s Theorem: A method of incorporating data into a prior distribution to obtain a posterior distribution. Bayesian Analysis: An approach that incorporates a prior distribution to express uncertainty about population parameter values. Best-Evidence Synthesis: Research syntheses that rely on study results that are the best available, not an a priori or idealized standard of evidence. Between-Study Moderators: Third variables or sets of variables that affect the direction or magnitude of relations between other variables. Identified on a between-studies basis when some reviewed studies represent one level of the moderator and other studies represent other levels. See Within-Study Moderators. Between-Study Predictors: Measured characteristics of studies hypothesized to affect true effect sizes. Between-Study Sample Size: The number of studies in a meta-analysis. 573 574 GLOSSARY Bibliographic Database: A machine-readable file consisting at minimum of entries sufficient to identify documents. Additional information may include subject indexing, abstracts , sponsoring institutions, and so on. Typically can be searched on a variety of fields (for example, author’s name, keywords in title, year of publication, journal name). See also Reference Database. Bibliographic Search: An exploration of literature to find reports relevant to the research topic. A search is typically conducted by consulting sources such as paper indexes, reference lists of relevant documents, contents of relevant journals /books, and electronic bibliographic databases. Birge Ratio: The ratio of certain chi-square statistics (used in tests for heterogeneity of effects, model fit, or that variance components are zero) to their degrees of freedom. Used as an index of heterogeneity or lack of fit. Has an expected value of one under correct model specification or when the between-studies variance component is zero. Buck’s Method: A method for missing data imputation that replaces missing observations with the predicted value from a regression of the missing variable on the completely observed variable; first suggested by Buck (1960). Categorical (Grouping) Variable: A variable that can take on a finite number of values used to define groups of studies. Certainty: The confidence with which the scientific community accepts research conclusions, reflecting the truth value accorded to conclusions. Citation Search: A literature search in which documents are identified based on their being cited by other documents. Classical Analysis: A analysis that assumes that parameter values are fixed, unknown constants and that the data contain all the information about these parameters. Coder: A person who reads and extracts information from research reports. Coder Reliability: The equivalence...

Share