Box Plot Calculator

Compare the center and spread of your data groups. Inspect quartiles, whiskers and flagged values, and see how each calculation connects to your observations.

Distribution explorer

See what the summary leaves out.

Compare location, spread, repeated values and fence flags. Keep the underlying observations visible, and choose a quartile convention explicitly.

3 of 3 groups available. Route A: median 10.5, interquartile range 3.5, 1 flagged observation.

Build your comparison

What are you measuring?

Methods can give different quartiles and fence flags. All groups use the same method.

Use the same quantity and unit for every group. Changing this label does not convert your data.

Your data groups

3 of 6

Paste observations or a value/frequency table. Each source keeps its own draft. Use the copy action to convert a valid dataset without changing its observations.

Route A dataset

Route A

Separate values with commas, semicolons, spaces or line breaks. No thousands separators or missing entries. Up to 2,000 observations.

Rename group
Route B dataset

Route B

Separate values with commas, semicolons, spaces or line breaks. No thousands separators or missing entries. Up to 2,000 observations.

Rename group
Route C dataset

Route C

Separate values with commas, semicolons, spaces or line breaks. No thousands separators or missing entries. Up to 2,000 observations.

Rename group

Up to 6,000 active input rows across groups. Values support signed decimals and scientific notation, up to 12 decimal places and magnitude 1 trillion.

Explore the comparison

Explore the data

Compare the distributions

Boxes show quartiles; the center line is the median. Whiskers reach observed values within the inner fences.

Shared scale · Delivery time (minutes)

Route A box plotMedian, lower and upper quartiles, and observed inlier whiskers on the shared scale. Exact values are in the summary, method details and data table.

Median 10.5 · IQR 3.5

Route B box plotMedian, lower and upper quartiles, and observed inlier whiskers on the shared scale. Exact values are in the summary, method details and data table.

Median 13.5 · IQR 2.5

Route C box plotMedian, lower and upper quartiles, and observed inlier whiskers on the shared scale. Exact values are in the summary, method details and data table.

Median 9 · IQR 2

5.8531.15

Delivery time (minutes)

One mark per distinct value: circles are inside the inner fences, diamonds are beyond an inner fence, and crosses are beyond an outer fence. Frequencies are shown in the inspector. Nearby marks may overlap at this scale.

30minutes

Arrow keys move between distinct values. Home and End reach the endpoints.

Frequency
1
Sorted rank interval
10–10
Strictly below this value
9 / 10
At or below this value
10 / 10

Beyond the upper outer fence

Route A

Median · minutes

10.5

The box spans 9.25 to 12.75 under Linear interpolation · R7.

Interquartile range
3.5
Observations
10
Flagged observations
1 of 10
Distinct values
8

Flags describe distance from the selected quartiles. They do not establish an error, statistical significance, or a reason to remove an observation.

Quartiles, fences & method comparison

Sorted ranks include every repeated observation. Interpolated quartiles may lie between observed values; they are different from the steps in the empirical cumulative distribution.

Lower quartile · 9.25

Between sorted ranks 3 and 4: 9 and 10. The upper value has weight 1/4; the remainder comes from the lower value.

Median · 10.5

Between sorted ranks 5 and 6: 10 and 11. The upper value has weight 1/2; the remainder comes from the lower value.

Upper quartile · 12.75

Between sorted ranks 7 and 8: 12 and 13. The upper value has weight 3/4; the remainder comes from the lower value.

Route A: the same observations under each quartile convention. Scroll for all columns.
MethodQ1MedianQ3IQRFlagged
Linear interpolation · R7 · selected9.2510.512.753.51
Tukey hinges910.51341
Median of halves · exclude middle910.51341
Observed minimum
8
Observed maximum
30
Lower whisker
8
Upper whisker
14
Lower inner fence
4
Upper inner fence
18
Lower outer fence
-1.25
Upper outer fence
23.25

Inner fences are Q1 − 1.5 × IQR and Q3 + 1.5 × IQR; outer fences use 3 × IQR. Equality with an inner fence stays inside. Outside the inner fences, equality at an outer fence is mild; only values strictly beyond an outer fence are extreme. Whiskers end at observed values within the inner fences.

Complete frequency and rank table
Route A: one row per distinct value; frequencies include all observations. Scroll for all columns.
ValueFrequencySorted ranksAt or belowFence classificationInspect
811–11 / 10Within the inner fences
922–33 / 10Within the inner fences
1024–55 / 10Within the inner fences
1116–66 / 10Within the inner fences
1217–77 / 10Within the inner fences
1318–88 / 10Within the inner fences
1419–99 / 10Within the inner fences
30110–1010 / 10Beyond the upper outer fence

Save and reuse your comparison

The report includes exact statistics, every distinct value, frequency, quartile convention and chart coordinate. Complete all active groups before exporting it. The input file can preserve unfinished drafts.

Selectable report CSV

Prepare the report for the current valid inputs. Editing a group or method clears stale report text.

Save or import a complete comparison

This complete-plan format preserves labels, methods and both data drafts. A malformed file leaves the current comparison unchanged.

Method, assumptions & sources

These are descriptive summaries of the entered observations. Frequency counts mean repeated observations, not bins or analytic weights. No value is automatically removed; box-plot overlap does not establish whether groups differ statistically.

Values are parsed as exact decimals. Quartiles, fences, ranks and counts retain exact values; only normalized chart coordinates use floating-point arithmetic. At a narrow scale, distinct marks may overlap. The inspector and tables retain their exact values.

R7 interpolates at rank 1 + (n − 1)p. Tukey hinges and median-of-halves use their stated conventions. For one observation, this tool defines all quartiles as that value, including in exclude-middle mode. Empirical cumulative shares count observations at or below an observed value and do not interpolate between values.

  • These are illustrative descriptive comparisons, not significance tests or population inferences.
  • Frequencies are positive integer multiplicities. Repeated observations are counted without expanding large frequency tables.
  • R7, Tukey hinges and exclude-middle halves can give different quartiles; all three use the same entered observations.
  • For exclude-middle halves with one observation, all three quartiles equal that observation.
  • Whiskers end at observed values inside or on the 1.5-IQR fences. Values strictly beyond 3-IQR fences are extreme; flags never remove data.
  • The empirical cumulative distribution is the fraction at or below an observed value. Vertical connectors depict jumps, not interpolated probabilities.
  • Measurement and unit labels are descriptive only; changing them does not convert observations.
  • Statistics, fences and counts are exact for the entered decimals. Normalized chart coordinates are approximate and nearby marks may overlap.

Method references: R sample quantiles, R box-plot hinges and whiskers, R empirical distributions, and NIST box plots. The input limits, strict missing-value policy and chart reference window are disclosed choices of this tool.

How to use this calculator

  1. Choose Quartile method, then enter Measurement label and Unit. The unit is descriptive text and does not convert observations.
  2. Enter a Group name for each of one to six groups. Choose Data source and enter either Raw observations or Value and frequency rows. Frequencies are positive integer counts of repeated observations, not weights or bin widths.
  3. Switch Data source to revisit its retained draft, or use an explicit copy action to convert the current data. An unsuccessful copy leaves both drafts unchanged. Correct invalid groups while continuing to inspect valid peers.
  4. Compare the aligned box plots and cumulative distributions. Choose Inspect group and Inspect distinct value to see the original value, its multiplicity, rank interval and cumulative counts. Whiskers are observed values, not the numeric fences.
  5. Review the three quartile conventions, exact summaries and flagged observations. Flags describe the entered sample; they neither delete values nor establish a statistically significant difference between groups.
  6. Use Input format and Complete comparison input to replace the whole plan atomically with CSV or TSV. Copy the reusable plan separately from Complete report CSV. The complete report is available only when every group is valid, and its selectable text matches the download.

The Formula

The mathematical logic powering this calculation is:

Sort observations with repeated values counted by their multiplicities. For R7, h = 1 + (n − 1)p at p = 1/4, 1/2 and 3/4; interpolate exactly between the surrounding one-based ranks.
Tukey hinges use lower rank floor((n + 3)/2)/2, median rank (n + 1)/2 and upper rank n + 1 minus the lower rank. Exclude-middle halves uses the median of each half, omitting the overall middle observation when n is odd. A singleton has all quartiles equal to its sole value under every offered method.
IQR = Q3 − Q1. Inner fences = Q1 − 1.5 IQR and Q3 + 1.5 IQR; outer fences = Q1 − 3 IQR and Q3 + 3 IQR. Whiskers end at the most extreme observed values inside or exactly on the inner fences.
A value outside an inner fence is mild unless it is strictly beyond an outer fence, when it is extreme. Equality with an inner fence is inside. Among values outside the inner fences, equality with an outer fence is mild. Flags never remove observations. Minimum and maximum remain separate from whiskers.
At observed value x, the empirical cumulative fraction is count(values ≤ x) / n; the jump size is frequency(x) / n. All groups share one linear measurement axis. Statistical quantities are exact for the entered decimals; normalized drawing coordinates are approximations, not measurement accuracy.

Worked Example

If you input:

  • Quartile method: Linear interpolation · R7
  • Measurement label: Delivery time
  • Unit: minutes
  • Route A — Raw observations: 8,9,9,10,10,11,12,13,14,30
  • Route B — Raw observations: 11,12,12,13,13,14,14,15,16,17
  • Route C — Raw observations: 7,8,8,8,9,9,10,10,11,12

Result:

Each route has 10 observations. With R7, Route A has Q1 9.25, median 10.5, Q3 12.75 and IQR 3.5 minutes. Its inner fences are 4 and 18, its whiskers end at 8 and 14, and 30 is flagged extreme but retained. Route B's median is 13.5 and Route C's is 9 minutes. These descriptive comparisons do not establish a statistically significant difference.

Frequently Asked Questions

Why can another calculator show different quartiles?

Sample quartiles have several conventions. R7 interpolates using rank 1 + (n − 1)p; Tukey hinges follow the R fivenum convention; Exclude-middle halves omits an odd sample's middle observation before taking the two half-medians. The selected method controls the box and flags, while the comparison table shows all three for the same data. R documents its quantile algorithms at https://stat.ethz.ch/R-manual/R-devel/library/stats/html/quantile.html and hinges at https://stat.ethz.ch/R-manual/R-devel/library/stats/html/fivenum.html. This tool does not claim complete compatibility with other software's parsing or missing-data behavior.

Are the whiskers the minimum and maximum?

Only when the extrema fall within the inner fences. Here whiskers end at the smallest and largest observed values inside or exactly on Q1 − 1.5 IQR and Q3 + 1.5 IQR. A numeric fence need not be an observed value, so fences, whiskers and full extrema are reported separately. The inclusive behavior follows the positive-coefficient whisker rule in https://stat.ethz.ch/R-manual/R-devel/library/grDevices/html/boxplot.stats.html; this workspace does not implement notches.

What do mild and extreme flags mean?

They are descriptive distance rules, not proof of error or a reason to delete data. Outside an inner fence is mild through and including the corresponding outer fence; strictly beyond Q1 − 3 IQR or Q3 + 3 IQR is extreme. Every observation remains in the quartile calculation and cumulative counts. NIST discusses inner and outer fences at https://www.itl.nist.gov/div898/handbook/eda/section3/boxplot.htm. The equality conventions stated here are deliberate; do not infer that every source or calculator uses the same rule.

What does the cumulative distribution show?

At a value, it shows the fraction of observations at or below that value. Repeated observations produce a jump equal to their count divided by the group's total; inspection also reports the count strictly below and above. Vertical connectors depict these jumps, not intermediate observed probabilities. See https://stat.ethz.ch/R-manual/R-devel/library/stats/html/ecdf.html. Unlike software that silently omits missing values, this tool rejects malformed active data as a whole group.

How do frequency tables and retained drafts work?

A frequency is the positive integer number of copies of an exact value. Equal values from multiple rows are combined; no expanded billion-row dataset is created. Raw observations and frequency rows keep separate drafts, so switching the source does not overwrite either. An explicit copy aggregates raw values or expands frequencies only when the destination limits allow it. Bounded unfinished inactive text is retained; a bad active group keeps valid peers visible but disables the complete report.

What happens with one value, repeated values or a zero IQR?

A single observation has equal minimum, quartiles, median and maximum; the singleton Exclude-middle rule is an explicit product convention. With IQR zero, the inner and outer fences coincide, so any unequal value is extreme under the stated rule. If every valid observation is equal, the chart labels its padded axis as a reference window rather than suggesting a nonzero spread. Narrow differences at large offsets are calculated before conversion to drawing coordinates. Small samples, overlapping boxes or equal medians do not establish equivalence, significance or causality.

Which input limits and formats are supported?

Each group accepts up to 2,000 raw observations or 500 frequency rows, with total frequency at most 1 billion. Across valid groups the active-row cap is 6,000. Values use decimal or scientific notation, at most 48 characters, exponent magnitude at most 24, absolute value at most 10¹² and at most 12 effective fractional places. Raw data accepts whitespace, comma or semicolon separators but rejects empty comma/semicolon cells. Frequency lines contain exactly value and count, optionally headed value,frequency or value,count; counts are unsigned digits from 1 to 1 billion. Quotes, expressions, NaN and infinity are not numeric values. The full-plan CSV/TSV format preserves both drafts and labels, including unfinished bounded data, and is distinct from the numerical report.

Similar Calculators