Skip to contents
metawoRld::launch()

opens the metawoRld app in your browser. It is a front door to the whole review for people who do not want to write R. Everything it does is done with the package’s functions, so the files it writes are the same, and any step can also be done (and reproduced) from R.

Overview, inbox and rejecting a submission whose value differs from the paper
Overview, inbox and rejecting a submission whose value differs from the paper

Start

Start a review with a short form: the title and question, what you compare (two groups, one group, or decide later), the main outcome and its unit, the criteria, and who you are. This calls new_review(), which sets up the group categories, the variable and a first meta-analysis, so the first study can be added right away. You can also open an existing project, or open the worked example to explore.

Starting a new review
Starting a new review

Overview

Progress at a glance (studies accepted, submissions waiting, values verified, the project check) and a list of next steps, each linking to the screen where it is done.

Inbox

Submissions from every channel arrive here: the entry form, CSV files, DataFindR, colleagues (through Sync). Each shows its groups, data points and the checks from _rules.yml. Accept or reject (with a reason) in one click.

When a submission is for a study that is already accepted, the Inbox shows exactly what would change (compare_with_accepted()), so a re-extraction cannot silently replace a correct value:

A re-extraction that would change one value
A re-extraction that would change one value

When two people extracted the same study, Compare and reconcile lists the disagreements; pick the right value for each and the app calls reconcile(), recording who agreed with every accepted value.

Review

The value review app (review_values()) as a tab: each value in plain language next to the paper, verified with the keyboard.

Reviewing values
Reviewing values

Setup

The parts of a project people actually change, without editing YAML: people and roles, variables (unit and other names), group categories (including rules over group facts), planned analyses (pick a variable, two groups and a measure), the criteria, and the form (shown as a readable list; add a group fact such as onset with add_group_fact()).

Planning an analysis
Planning an analysis

Check and Publish

Check runs check_project() and lists problems in plain language; Publish builds the living meta-analysis website with Quarto and opens it.

Working as a team: Sync

Projects can keep their history in Git, and the app hides the Git commands:

  • After each action (accept, reject, reconcile, a change in Setup) the app saves a commit with a readable message (“Accept gencheva2022 (…) by arb”), so the history reads like a log of the review.
  • Sync saves your work, fetches the team’s work and sends yours (sync_project()). New submissions from colleagues then appear in the Inbox.
  • Conflicts are rare by design: each extraction has its own folder in submissions/, only arbiters write data/, and screening.csv (which several people append to) is merged line by line.
  • The app uses the system git command, so your existing sign-in (a credential manager, GitHub Desktop, SSH keys) works. To share a project, create an empty repository (e.g. on GitHub) and paste its address in the Sync dialog. Without write access, push to a fork and open a pull request.