Does each dot always represent one observation?

Usually in an observation-level dot plot, but any different convention must be explicitly stated.

Why are dots stacked?

Stacking prevents repeated or nearby values from hiding behind one another and makes their frequency visible.

Can a dot plot show a mean?

A summary marker can be added, but it should be visually distinguished from dots representing individual observations.

AI Glossary / Statistics and probability / Beginner

Dot plot

Definition

A dot plot, in the distribution-display sense used by the source glossary, represents observations as dots positioned along a measurement axis, with repeated or nearby values stacked. Each dot usually represents one observation, making frequencies visible without replacing them entirely with bars. It is especially useful for small datasets. The name also refers to other chart types, including Cleveland dot charts for category-level values, so the plotted quantities and dot meaning should always be labeled.

What it means

The display helps readers see where observations concentrate, how far they spread, whether gaps occur and which values are unusual relative to the rest. A stack of three dots means three represented observations at that value or in that bin, not a measurement value of three on a second scientific variable. Compared with a single mean or bar showing a group average, an observation-level dot plot reveals individual variation. Its interpretation depends on whether dots preserve exact positions or use bins, rounding or jitter.

How it works

Choose the numerical variable, units and observations to display. Place values on a labeled horizontal or vertical axis. Stack dots at identical values, or use clearly specified bins when nearby observations would overlap. Keep dot sizes consistent and state whether one dot means one observation or another count. If comparing groups, use compatible scales and explain whether dots show individuals or summaries. For dense data, consider whether another display is clearer. A dot plot is descriptive: judging uncertainty, dependence or statistical significance requires additional analysis rather than reading stack heights as a test result.

Three practical examples

Service review: show individual task durations for a small pilot so managers can see variability beyond the average.
AI evaluation: display per-run latency or performance measurements while identifying how repeated runs were generated.
Training: visualize frequencies in a short numerical dataset and connect stacks of dots with counts and proportions.

Think of it this way

Think of placing one token for every observation on a numbered ruler. When several tokens belong at the same position, stack them so none disappears behind another. The ruler shows the measured value; the stack tells you how many observations share that location. Neither alone explains why the values occurred.

Worked example

Suppose eight hypothetical waiting times are 2, 2, 3, 3, 3, 5, 8 and 8 minutes. A dot plot stacks two dots at two, three at three, one at five and two at eight. The four stack counts sum to eight, confirming that all observations are represented. The most frequent value is three minutes, occurring in 3/8 = 37.5% of observations. The median is three minutes, while the mean is 34/8 = 4.25 minutes. Two observations exceed six minutes, giving an observed proportion of 25%. There are no observations at four, six or seven in this small dataset, but those empty positions do not prove the population cannot produce those waits. Plotting only the mean at 4.25 would conceal the repeated values and the two longer waits.

Illustrative case study

Fictional case: Rowan Workflow runs a small comparison of service procedures and initially presents only average completion times. An analyst adds observation-level dot plots and notes that repeated measurements from the same operator are not independent new operators. The team labels groups consistently and retains the raw values for further analysis. Managers use the display to identify questions about unusually long tasks rather than immediately declaring a winning procedure. This invented case illustrates transparent descriptive reporting, not evidence of a statistically reliable improvement.

Common misconceptions and mistakes

Do not confuse a stacked distribution dot plot with a scatterplot of two measured variables or a Cleveland chart of category summaries. Binning can hide distinctions between nearby values, and jitter moves points for readability rather than creating new measurements. Unequal dot sizes can imply unintended weights. A larger sample naturally creates more dots, so comparisons of stack heights need context. Visual gaps or isolated points in a small sample are not definitive population features. Showing every observation also does not make a nonrepresentative sample representative or reveal dependence automatically.

Business use

Businesses can use dot plots to make small datasets accessible and to show the observations behind pilot results. They encourage discussion of variation, unusual cases and sample size. Label the measurement, grouping and dot unit, and include uncertainty summaries only when their calculation is appropriate. For large monitoring datasets, a histogram, empirical cumulative curve or another display may communicate more clearly. The best choice follows the question and audience rather than a rule that one chart type is always superior.

Check your understanding

Why is the mean in the example higher than the median?

Answer: The two eight-minute waits pull the arithmetic mean upward, while the middle two observations in the ordered list are both three. The dot plot makes this asymmetry visible without requiring readers to infer it from a single average.

Related terms

Histogram
Empirical distribution
Median
Mean
Exploratory data analysis
strip plot
stacked dot plot
Cleveland dot chart
frequency
jitter
binning

Learn next

draw a dot plot from a short ordered list and verify the total number of dots. Compare it with a histogram and a plot of the mean alone. Then explore how binning, jitter and grouping change readability, while keeping the distinction between individual observations and summary statistics clear.

Sources

NIST/SEMATECH, glossary, dot plot: https://www.itl.nist.gov/div898/handbook/glossary.htm Wilkinson, Stacking Graphic Elements to Avoid Over-Plotting: https://www.cs.uic.edu/~wilkinson/Publications/stacking.pdf R graphics documentation, Cleveland's Dot Plots, alternative terminology: https://search.r-project.org/R/refmans/graphics/html/dotchart.html

Previous: Dormand - Prince method Next: Dot product