Choose a form for the question
Working examples, data contracts, and the choices that keep a chart honest.
Ranking under uncertainty
Expose the uncertainty a league table usually hides.
Confidence ribbon
Show estimation uncertainty across a sequence.
Marginal effect
Compare a raw association with a model prediction holding covariates fixed.
Forecast fan
Show a predictive distribution instead of a single promised future.
Caterpillar plot
Order a dense set of estimates while retaining their intervals.
Transition matrix
A compact view of where observations move between two ordered states.
Value heatmap
Compare a quantity across two categorical dimensions, without hiding the numbers.
Scatter with binned summary
Keep the observations visible while revealing their local average relationship.
Faceted comparison
Repeat one chart across groups so differences can be read on common scales.
Forest plot & ranked estimates
Show an estimate and its uncertainty on a shared quantitative axis.
PCA biplot
Show observation scores and variable directions in the same reduced-dimensional view.
Scatter with empirical envelope
Show how a distribution widens or narrows along another variable.
Diverging distribution
Compare ordered response distributions around a declared central category.
Rate with exposure
Read a rate across bins together with how many observations each bin holds.
Difference lines
Show two series and the gap between them, repeated across groups on one scale.
Interrupted time series with a movable cut
Test a dated break by moving the cut and watching where the estimated jump is largest.
Heatmap timeline
Many rows of one indexed quantity over time, with a brushable overview that zooms the grid.
Synced maps
Two or more maps of the same place, side by side on one extent, read through one cursor — for then-and-now comparisons place by place.
Ridgeline (joyplot)
Stack one distribution per group, in order, so a shift in the whole shape reads as movement down the page.
Concentration (Lorenz) curve
Show how much of an outcome the most-concentrated share of a population holds.
Circumplex wheel
Place estimates where a circular theory puts them, so the reader can see whether the data respect the circle.
Interval scatter
Place units by two estimates at once, each with its own interval, so both kinds of uncertainty stay in view.
Linked map and scatter
One set of places drawn twice — where they are, and what they are — so a pattern found in either view can be located in the other.