dv-dashboard

Interactive dashboard showcasing trends and summary statistics for the datasets in the CAFE Dataverse collection and its subcollections.

Live at https://climate-cafe.github.io/dv-dashboard

The dashboard summarises the 1,307 published datasets in the CAFE Research Coordinating Center collection on Harvard Dataverse: what the collection holds, who contributed it, what it covers in space and time, how the subcollections nest, and which datasets resemble each other by keyword and by description.

Where things are

Path	What it is
dashboard-site/	Everything: the analysis pipeline and the site
dashboard-site/README.md	Start here
dashboard-site/analysis/README.md	The audit guide: every script, what it decides
dashboard-site/deploy/README.md	Publishing to GitHub Pages
setup/SETUP.md	The original brief
dv-data/	The SQL extract set (gitignored)

Building it

cd dashboard-site
just setup     # restore the R (renv) and Python (uv) environments
just fetch     # re-acquire public metadata from Harvard Dataverse
just build     # run the pipeline: R descriptives, then Python similarity
just serve     # preview at http://localhost:8000
just check     # lint, tests, and the privacy scan

just build needs the extract set at dv-data/. Without it, the committed site data still serves and just serve still works.

How it is organised

All data processing is in scripts that run on their own. R does the descriptive layer and Python does the semantic-similarity layer, so the work is reviewable by both halves of the team, and any single script can be run and audited in an interactive session without the dashboard running. The site's JavaScript filters, sorts, and draws; it computes no statistics of its own, and a test asserts that its unfiltered totals match the ones R produced.

Two things to know before reading a number

Drafts are excluded, as the brief requires: 545 of the 1,852 datasets in CAFE scope.

"Published" is not "visible." Harvested datasets are fully live but carry no publication date, so filtering on publication date is not null silently drops 513 of 1,307 datasets. The test used throughout is published locally OR harvested. The dashboard's Methods page covers this and every other decision that moves a number.

Data

dv-data/ is gitignored: how those extracts will be shared has not been settled, and publication to Dataverse is one option under consideration. The aggregated payload the site draws from is committed under dashboard-site/public/data/, and carries no personal data. See dashboard-site/deploy/README.md for what that means for deployment, and the dashboard's Methods page for the privacy decisions behind it.