LoginSign Up
← Back to all sparks
G

ggcorrplot

ANALYTICS
Velocity2.5

Visualization of a correlation matrix using ggplot2

github.com/kassambara/ggcorrplot ↗

ggcorrplot came back after four years and found its significance markers had been lying

correlationrggplot2visualizationsignificance testingbug fixes
◆Current state
ggcorrplot draws correlation matrices in ggplot2 with optional significance marking and hierarchical reordering. It sat untouched from late 2022 until mid-2026, then shipped 0.2.0 and 0.3.0 sixteen days apart. Between them they added the display options users had been requesting since 2016 and repaired a set of bugs where hc.order = TRUE silently changed which cells were marked significant.
◆Where it's heading
Both releases chase the same target: parity with the older corrplot package inside a ggplot2 object. Significance stars appended to coefficient labels, circle scaling, decimal control, then boxed cells and glyphs sized by absolute correlation — these are corrplot's visual vocabulary reimplemented where they can be composed with other ggplot2 layers. The bug fixes point the other way, at foundations: p-values matched to cells by name rather than row position, clustering computed on the unrounded matrix, tl.col actually applied.
◆Prediction
With the corrplot look largely reproduced and the correctness backlog cleared, the remaining gap is the mixed upper/lower display corrplot supports; that is the natural next argument if the current release pace holds.

◆Recent moves

  1. 21d ago

    ggcorrplot 0.3.0 adds boxed cells and correlation-sized squares

    Visual parity work continued from 0.2.0: cell.grid draws a light rectangle behind each glyph and drops the through-center gridlines for corrplot's boxed-cell look, and scale.square sizes squares by absolute correlation the way circles already were. Both default off, so existing plots are unchanged.

    View source ↗
  2. 1mo ago

    ggcorrplot 0.2.0 fixes significance markers broken by hc.order

    The revival release, and it is mostly repair. sig.stars, circle.scale and nsmall are the visible additions, but the substance is a set of fixes where hc.order = TRUE rounded the p-value matrix before comparing it to sig.level — so a p of 0.054 could be marked significant with reordering on and not without. Clustering also ran on the rounded matrix, and markers were matched to cells by row position rather than name.

    View source ↗
  3. 3y ago

    ggcorrplot 0.1.4

    The last release before dormancy: an as.is argument, visual regression testing via vdiffr, and fixes making hc.method and show.diag actually take effect. The vdiffr infrastructure is what made the 2026 rendering changes safe to attempt.

    View source ↗
  4. 6y ago

    ggcorrplot 0.1.3

    Support for cor_mat objects from rstatix, tying ggcorrplot to its sibling package's test output. Otherwise style and spelling cleanup from a contributor pass.

    View source ↗
  5. 7y ago

    ggcorrplot 0.1.2

    Two small arguments — digits and ggtheme — plus label and package-collision bug fixes. Routine early-life maintenance.

    View source ↗
  6. 10y ago

    ggcorrplot's first release: correlograms in ggplot2

    The original release, offering ggcorrplot() for visualizing a correlation matrix with reordering and significance levels, plus cor_pmat() for the p-value matrix. The p-value handling introduced here is what 0.2.0 spent a decade later correcting.

    View source ↗