Loading the Hagen Cumulative Science Project…
This can take a few seconds — the app has to load its full study database.
Drag the title bar to move the window; click a band to filter the whole app by that range.
Each point is one study; point size reflects the replication's statistical power. The dashed line marks a perfect replication (equal r). Click a point to open that study in the Study Detail tab. Original → replication moves every study from the dashed line to its replication effect size (the distance travelled is the shrinkage).
Each bar is a topic. Studies whose titles and abstracts use similar words were grouped into topics (at least six studies each) (OpenAlex texts). A topic is named after three of its typical words, so the name is a hint, not a definition. The bar shows the share of successful and failed replications among the studies currently shown, n = number of studies.
One row pair per study: original study (black) and replication (blue = successfully replicated, orange = failed to replicate). Only studies matching the filters of the All Studies tab are shown. Click a point to open that study in the Study Detail tab.
† undirected model: the CI is a magnitude only, so it can never cross zero or point the other way; the sign shown is attached afterwards (see the footnote in the All Studies tab). "CI incl. original" = the replication's 95% CI contains the original effect size.
Click a row to open that study in the Study Detail tab.
† undirected model: the effect involves more than one comparison at once (e.g. a factor with 3+ levels, or an r2beta-based mixed model), so its CI (square-rooted eta squared / partial R2) is a magnitude only. The sign shown is added afterwards by hand, based on the comparison the study's hypothesis is actually about (e.g. "condition A > condition B"), just so it can be plotted on the same scale as the other, directed studies -- unlike those, this CI can never cross zero or point the other way.