Describe › Standalone plots

Estimation / Gardner-Altman plot

An estimation plot (Cumming / Gardner-Altman) shows jittered raw data for two groups beside the mean difference and its 95% bootstrap CI — effect-size focused, with no p-value.

What is Estimation / Gardner-Altman plot?

Estimation plots are the 'new statistics' alternative to t-test bar charts. The framing: show the raw data + the effect size + its uncertainty, not the (often misleading) significance test. Cumming's argument is that a CI on Δ is what readers actually need to decide whether the effect matters.

Δ axis is anchored at the mean of group 1 by convention; a dashed reference line extends from group 1's mean to the Δ position. The 95% CI is from a percentile bootstrap (default 1000 reps) on the difference of means.

When should I use Estimation / Gardner-Altman plot?

  • Two-group comparisons where effect size matters more than significance.
  • Modern visualisation standard in fields adopting the 'new statistics' framing (psychology, ecology).

What data does it need?

Numeric value + grouping column (exactly 2 levels) + bootstrap reps.

What does it report?

Combined dot plot + Δ marker + 95% bootstrap CI + summary of means and CI on Δ.

How do I interpret the result?

The CI tells you the plausible range of Δ; the width tells you how precisely you've estimated it. Width as informative as point estimate.

See also

References

  • Cumming (2014). The new statistics: why and how. Psychological Science 25(1).