The main page of the application displays the number of papers, studies, effect sizes, and participants contained in the databank, and allows the user to visualize these studies according to the country where the studies were conducted, sample size per study, and year of publication. Users can conduct a search to select studies that contain effect sizes on specific topics, such as personality, communication, incentives, framing, and reputation. The selection of studies can be refined based on more specific criteria, such as the type of personality trait (e.g., Social Value Orientation or Honesty-Humility), communication occurrence (e.g., one-shot or ongoing), and type of reputation manipulation (e.g., anonymity or gossip). The platform allows users to also select studies based on specific sample and study characteristics. Users can personalize the meta-analysis by removing individual studies. The selection of studies and their study characteristics are displayed in a table, and can be easily downloaded. Users can also efficiently obtain a list of references of the selected studies that can be imported into a reference manager software. This tutorial will describe these functions of CoDa.
After the user completes a selection of studies, the user can obtain the output of the meta-analytic results by selecting the “meta-analyses” tab. The calculations are conducted using R’s package metafor. Users can select between options about (1) the effect size used in the meta-analysis (e.g., r, d), (2) random versus fixed effects models, and (3) estimators of residual heterogeneity. The output of the meta-analysis is displayed in a table and the raw R output is also obtained. The standard output includes an overall effect size estimate, 95% confidence interval, 95% prediction interval, and estimates of heterogeneity. Finally, the user may also conduct multilevel meta-analysis to account for dependencies in the data. For example, it is possible to specify a model that includes a variable for papers (or countries) as random effects if the meta-analysis includes multiple studies from the same paper (or the same country). This tutorial will breifly describe these analyses and output.
Users may conduct meta-analyses on standard effect sizes representing the relation between two variables by selecting the “meta-analyses” tab, but can also meta-analyze standardized mean levels of cooperation observed across studies and/or treatments (i.e., meta-regression) by selecting the “meta-regression” tab. Meta-regression can be conducted for any set of treatments that are specified by users in Treatment #1 of the selection tool. Users can select between options about (1) random versus fixed effects models and (2) estimators of residual heterogeneity. The output of the meta-regression is displayed in a table and the raw R output is also obtained. The standard output includes an overall effect size estimate (logged rates of cooperation), 95% confidence interval, 95% prediction interval, and estimates of heterogeneity. Finally, the user may also conduct multilevel meta-regression to account for dependencies in the data. For example, it is possible to specify a model that includes a variable for papers (and/or studies, countries) as random effects if the meta-analysis includes multiple studies from the same paper (or the same study/country). Users can also select moderators to predict the logged rate of cooperation. This tutorial will describe how to use meta-regression analyses in CoDa.
Specific research questions can suffer from publication bias, which can result from (a) statistical significance being a criterion for publication and/or (b) questionable research practices (e.g., p hacking, selective reporting). CoDa provides users an ability to estimate publication bias (i.e., Trim-and-Fill, Egger’s Regression, and Henmi-Copas). This tutorial will describe how to assess publication bias using these analyses reported in CoDa.
CoDa offers the ability to specify multi-level models for meta-analyses and meta-regression. You can indicate whether you want to nest effect sizes (or logged cooperation rates) within studies and papers. You are also able to specify country as another level, which can be crossed with papers (and/or studies). You can also select moderators to predict the effect size, or in the case of meta-regression logged cooperation rates. You can choose from several moderators, including moderators that are specific to the variable under investigation (e.g., punishment agent, punishment iterations), study characteristics (e.g., group size, proportion male), and country indicators (e.g., rule of law, trust). This tutorial will describe how these features work and provide some guidance in how to interpret results.
One major benefit of meta-analysis is that the estimate of a population level effect size can be used in statistical power analysis to calculate the sample size required in the next study that investigates a certain effect. While doing a meta-analysis using CoDa, the effect size estimate is used as input in a statistical power analysis to estimate the adequate sample size required in future studies to detect this effect size. The user can further customize these analyses by adjusting assumptions about the statistical test, and the levels of alpha and beta of these analyses. The power analyses are based on the pwr R package. This tutorial will describe how to use power analyses using CoDa.
The Citation Explorer displays the citation network of studies annotated in CoDa, enabling users to identify communities and topics within the history of research on human cooperation. The citation network models the papers included in the databank (for which a DOI could be found), and established links between the papers based on their references. The citations were obtained from CrossRef and Microsoft Academics, using the papers’ DOIs as entry points. Then, communities were identified based on their modularity, i.e., the density of the links intra- and inter-communities, using the Louvain community detection algorithm. The visualizations can be customized by users, to explore studies based on the information annotated in CoDa, This tutorial will briefly introduce and describe the citation explorer in CoDa.
CoDa offers an Ontology Explorer, which allows users to visualize and dynamically interact with the domain-specific schema of human cooperation studies. All classes are subclasses of the generic class “Independent Variable”, corresponding to the outer circle. The size of a circle represents the number of treatments annotated with that variable included in CoDa. The user can hover on a circle to display the definition of a desired class (e.g. Personality), and additionally zoom in and obtain the labels and definitions of its subclasses. The definitions and possible values of each Independent Variable is provided in a table. This tutorial will describe how to use the Ontology Explorer to learn more about what information is included in CoDa.