Sign in Contact us Team FAQ About Discover

Searching and Organization

  • How does the search work?

    The search window will display any models or meta-analyses with the terms you selected in their title, description, or categories. You can also filter by models and meta-analyses with specified dependent or independent variables.


    Searching models and meta-analyses.

    search, variables, filtering

  • What are categories?

    Categories are a way to organize content on the site, and work like a hierachy of tags. Every model and meta-analysis can be placed within multiple categories, and the categories are organized within eachother. Categories are used both for searching and for browsing.


    Hierarchy of tags.

    category, browse, search

Model Formats

  • What are models?

    Models are parameter estimates. It can be a Dose-Response model, a Treatment-Control model, a Marginal Effect Distribution, or any other model which specifies for a set of independent variables the distribution of a dependent variable.

    Models come from scientists and from scientific articles.

    Model Attributes

    Attributes for the published article and parameters of a model.

    model, distribution, estimate

  • What is the format for model information?

    Although there are a number of ways to specify models, the most general is to upload a model file. There are a number of specifications. See the Model Formats Specifications for more information.

    model, format, upload, download

Graph Displays

  • What kinds of graphs are available?

    Here are some of the available graphs on the site, and where they're used:

    Stacked Plot

    For discrete probability models.

    Bar Plot

    For categorical splines.

    Curve Plot

    For dose-response curves.

    Spline Distribution

    For a single parameter, or pooled across models.

    Sampled Distribution

    For a single parameter, or hierarchically simulated across models.

    models, graphs, plot

Meta-Analyses and Merging

  • What are meta-analyses?

    Meta-Analyses are sets of models that describe the same dependent variable. By organizing models into meta-analyses, you can merge together the results into combined best-estimate.

    meta-analysis, merge, meta-analysis

  • How can models be merged?

    Models can be merged to produced "Pooled" estimates or "Hierarchical Normal" estimates. A pooled estimate is appropriate if all of the models are estimating the same thing. Hierarchical Normal estimates assume a hierarchical structure, whereby each model is an instance of a more general model, with an unknown dispersion between the models. This allows for "partial pooling".

    • The merged result can only be estimated with simulations, so it will be uneven. Currently the plots are generated from 1000 sampled points, and plotted by bin counting.
    • Each partial pooling result requires an analytical model. The site currently supports Gaussian models and automatically approximates any model as Gaussian for that method

    meta-analysis, merge, pooling, bayesian

Scaffolds and Pattern Templates

  • What are scaffolds?

    Scaffolds are a way to simplify the model input process. Every meta-analysis has a scaffold, and you can customize it to specify which fields should be hidden, pre-entered, and required. Then you can send it out to others to collect information.

    scaffold, template, meta-analysis

Distributions and Maps

  • What are distributions?

    Distributions are a way to apply models to populations-- or, equivalently, to change the independent variable of a model.

    distribution, variable, population

Combinations and Comparisons

  • What are combinations?

    Combinations are linearly scaled sums of models, which are useful for generating combined results across different dependent variables. The same page can be used to compare two models or meta-analyses.

    combination, comparison, linear