How to Extract Data from a Box Plot Image (Five-Number Summary)
⚡ Quick Answer
Manually, in about two minutes per plot. Set the Y axis in Engauge Digitizer, then use the point tool to click the whisker ends (min/max), box edges (Q1/Q3), the median line, and any outlier dots. You recover the full five-number summary — enough to compare groups or feed a meta-analysis. Forest plots work the same way: read each study’s point estimate and CI whiskers.
Why Box Plots Need Manual Reading
Automatic curve tracing expects a continuous line. A box plot is a set of discrete landmarks — whiskers, box edges, a median tick — so the tracer has nothing to follow. The good news: those landmarks are exactly what you need, and clicking them in the zoom window is fast and precise.
Step-by-Step: Reading a Box Plot
- Set the Y axis: In the Axis Wizard, calibrate the value axis on two labeled gridlines. The X (group) axis doesn’t need calibration — you only need Y values.
- Read the five landmarks: With the point tool, zoom in and click: lower whisker end (minimum), box bottom (Q1), median line, box top (Q3), upper whisker end (maximum).
- Capture outliers: Click each plotted outlier dot as well — they matter for distribution shape and for spotting data-entry errors in the original.
- Label per group: Add one “curve” per group in the Curves List (“Control”, “Treatment”…) so the CSV stays organized. Export.
What You Can (and Can’t) Recover
- You get: min, Q1, median, Q3, max, outliers — the complete five-number summary.
- You can approximate: SD ≈ IQR / 1.35 for roughly normal data (IQR = Q3 − Q1). Label it an approximation.
- You can’t get: the raw data points, exact SD, or sample size — a box plot simply doesn’t contain them. Check the caption for n.
Forest Plots: The Same Skill, Meta-Analysis Payoff
Forest plots (one row per study: a square point estimate with CI whiskers) digitize identically: read the square and both whisker ends per row. Record study, estimate, lower CI, upper CI — those four columns are exactly what meta-analysis software wants. Pair this with our bar-chart error-bar guide and you can rebuild most published meta-analysis datasets from figures alone.
⚠️ Note: Box plots hide distribution shape — two very different datasets can share a five-number summary. Recovered quartiles are great for comparison, but don’t claim more than the plot shows.
Related Guides
How to Digitize Bar Charts with Error Bars
Means and uncertainty — the companion meta-analysis skill.
Is Digitized Data Accurate Enough to Publish?
Validation evidence for figure-extracted data.
How to Digitize Well Logs
Another niche chart type, same manual-reading skills.
Frequently Asked Questions
Can Engauge Digitizer automatically read a box plot?
No — box plots have no continuous curve for the auto-tracer to follow. The reliable method is manual: set the Y axis, then click the whisker ends, box edges, median line, and outlier dots with the point tool. A typical box plot takes about two minutes.
What statistics can I recover from a box plot image?
The five-number summary: minimum, first quartile (Q1), median, third quartile (Q3), and maximum — plus any plotted outliers. That is enough to compare groups or approximate the distribution shape.
Can I recover the standard deviation from a box plot?
Not exactly. A box plot doesn’t contain the SD, but you can approximate it: for roughly normal data, SD ≈ IQR / 1.35, where IQR = Q3 − Q1. Report it as an approximation, never as a measured SD.
How do I digitize a forest plot?
Same technique: set the value axis, then for each study row read the point estimate (the square) and both confidence-interval whisker ends. Record study name, estimate, lower CI, upper CI — exactly the inputs a meta-analysis needs.
Ready to try it yourself?
Download Engauge Digitizer free and follow along with this guide.
Download Engauge Digitizer