From a Graph to a Useful Description
A graph can show many details at once, but a reader usually needs a short account of the distribution’s main pattern. For a quantitative variable, a useful description reports its shape, typical location, variability, and any unusual features, all in the context of the variable being measured. These parts help someone understand the data without having to inspect every point or bar.
The goal is not to list everything visible. It is to select the most informative features and describe them accurately. A description should tell the reader what was measured, what the overall pattern looks like, where observations tend to lie, how much they vary, and whether there are notable gaps, clusters, or observations apart from the rest. It should not turn a visual pattern into an unsupported claim about why the pattern occurred.
A Practical Description Framework
A helpful way to organize a description is to move from the broad pattern to the details. First name the variable and its units. Then describe the shape, give an appropriate indication of typical location, summarize variability, and note any features that stand out. The details available depend on the graph: a histogram displays the distribution’s overall shape, while a boxplot displays quartiles and possible flagged observations but hides much of the distribution’s shape.
Say what each observation represents, such as a student’s commute time in minutes or a package’s mass in grams. This makes the description meaningful outside the graph.
Use terms supported by the display, such as roughly symmetric, skewed to the right, skewed to the left, or having distinct clusters. For a skewed distribution, identify the direction of the longer tail.
Give an approximate value or central interval that represents where observations tend to lie. If the graph shows distinct clusters, describe their locations rather than forcing them into one typical value.
Give an approximate span or another measure visibly supported by the graph. Note clear gaps, separated clusters, or observations far from the main body, using cautious language.
Descriptions should be calibrated to the display. A histogram with wide bins does not reveal every individual value, so its apparent endpoints and center are approximate. A boxplot can provide a median and quartiles but cannot show whether values within each quartile are clustered, spread evenly, or separated by a gap. State what the graph does show, and do not imply that it supplies information it hides.
Worked Example: Describing a Right-Skewed Histogram
Suppose a hypothetical histogram summarizes commute times for 60 employees at one office, measured in minutes. Its bars show that most times lie between 15 and 35 minutes, with the greatest concentration around 20 to 30 minutes. The bars become less frequent toward higher times; a few observations extend to about 55 minutes. The shortest time is about 5 minutes.
Shape. The distribution is right-skewed: its longer tail extends toward larger commute times. This describes the visible pattern; it does not explain why some commutes are longer.
Typical location. A reasonable visual summary is that commute times are typically around 20 to 30 minutes, where the histogram has its greatest concentration. Since the bins cover intervals rather than displaying every exact value, this is an approximate description rather than a precise center.
Variability. The plotted times extend from about 5 to about 55 minutes. Using the graph’s approximate endpoints, the span is about \(55-5=50\) minutes. Many observations are more concentrated, between 15 and 35 minutes, but the full distribution covers a broader range.
Unusual features and full description. A few longer commutes stretch the right tail. A concise description is: “For these employees, commute times are concentrated around 20 to 30 minutes and are right-skewed. Most are between about 15 and 35 minutes, although the observed times span roughly 5 to 55 minutes, with a few longer commutes.” The wording reports the pattern and its context without proposing a cause.
Location and Variability Must Fit the Graph
“Typical” does not always mean one particular number. When a graph has one clear mound, a visual estimate of its central region may communicate typical location well. If a boxplot is available, its median is a specific measure of center. In either case, include units and avoid giving more digits than the graph supports. A central estimate of 24.7 minutes would not be justified by a histogram whose scale and bins only support an estimate to the nearest several minutes.
Variability also has more than one useful description. The full span from the smallest to largest observation gives a broad sense of spread, but it is strongly affected by the endpoints. The width of the central portion may be more representative when the graph shows a long tail or an isolated value. If using quartiles from a boxplot, describe the interval directly—for example, “the middle half of the observations lies between about 12 and 20 minutes.” The interquartile range is the width of that interval, but a contextual sentence often makes the location of the interval clearer.
A description should not rely on a numerical summary alone. “The median is 16.5 minutes” gives a location, but says nothing about skew, spread, or unusual values. Conversely, “the distribution is right-skewed” gives shape but leaves the reader without a sense of the actual commute times. Several complementary details make the description informative.
Worked Example: Describing a Dotplot with a Separated Observation
A hypothetical dotplot gives the times, in minutes, for 10 students to complete a practice task: 12, 14, 15, 15, 16, 17, 18, 20, 24, and 41. The display shows most points between 12 and 24 minutes and one point at 41 minutes, separated from the rest.
Shape and location. Most task times form a concentration between 12 and 20 minutes, with the central observations near 16 to 18 minutes. The separated high value gives the distribution a longer right side. It is clearer to report the main cluster and the separated observation than to imply that all 10 times lie near one typical value.
Variability. The observed span is from 12 to 41 minutes, so its width is \(41-12=29\) minutes. The first nine observations lie between 12 and 24 minutes, while the final observation is 41 minutes. This makes clear both the main group’s spread and the contribution of the separated point.
Contextual description. “Most students completed the task in about 12 to 24 minutes, with a concentration around 16 to 18 minutes. The distribution has a longer right side because one student’s time, 41 minutes, lies apart from the rest; the overall times span 29 minutes.” Calling that observation “an error” would go beyond the evidence in the dotplot. It may deserve checking, but the graph alone cannot establish that it was recorded incorrectly.
Clusters, Gaps, and the Limits of Boxplots
A distribution may have two or more clusters rather than one central mound. In that case, naming only a single typical value can conceal an important feature. Describe where the clusters lie and whether there are gaps between them. A graph can show that observations form separate groups, but the visual separation alone does not identify what distinguishes the individuals in those groups.
Boxplots require particular care when describing shape. The median and quartiles locate the center and divide the ordered observations into four parts, but a boxplot does not show the detailed arrangement of values inside those parts. Similar boxplots can come from distributions with different shapes. Therefore, a boxplot can support statements about a median, quartiles, and spread between quartiles; it cannot, by itself, justify calling a distribution symmetric, skewed, or bimodal.
Worked Example: Describing Two Clusters in a Histogram
Consider a hypothetical histogram of the number of minutes 40 devices take to complete a task. The bars show one concentration from about 10 to 20 minutes and another from about 35 to 45 minutes, with few observations between about 22 and 32 minutes. The plotted values extend from about 8 to 48 minutes.
Shape and location. The histogram shows two clusters, one around 10 to 20 minutes and another around 35 to 45 minutes, with a gap-like region between them. Giving only a single typical time would hide this two-cluster pattern. The graph supports reporting both cluster locations, not explaining why the devices fall into them.
Variability. The approximate observed span is \(48-8=40\) minutes. The gap-like region is another important aspect of the distribution: it shows that values are not spread evenly across that span.
Contextual description. “Task-completion times for the 40 devices form two clusters, around 10 to 20 minutes and 35 to 45 minutes, with few times between about 22 and 32 minutes. The overall span is roughly 8 to 48 minutes.” This description includes shape, location, spread, and an unusual feature while leaving possible explanations open.
Why Context and Restraint Matter
A graph describes the observed distribution; it does not, by itself, explain how that distribution arose. A longer right tail is evidence of some larger observations, not evidence that a particular event caused them. Two clusters do not prove that the observations belong to two known types of individuals. Those explanations may be worth investigating with additional information, but they should not be included as established facts in a graph description.
Context also helps prevent ambiguous statements. “The values vary from 5 to 55” is less useful than “employee commute times vary from about 5 to 55 minutes.” A reader needs to know which quantity the numbers refer to. When the graph is based on a particular group, identify that group as well, so the description does not imply that the observed pattern necessarily applies to everyone.
A short justification for including both location and spread is that either one alone leaves out important information. Two distributions can have observations centered in a similar region but very different spread; two can have similar spread but be centered in different regions. A description that reports both therefore distinguishes patterns that a single summary cannot.
This is an organizing checklist, not a calculation or a guarantee that every part will be equally clear in every graph. If a feature cannot be read reliably, qualify the statement or leave it out rather than guessing.
Common Mistakes
- Giving shape without context. “It is skewed” does not tell the reader what is skewed. Name the variable, group, and units where appropriate.
- Reporting only a center. A single typical value can hide a long tail, wide spread, gap, or multiple clusters. Add the most relevant information about variability and shape.
- Claiming precision the graph does not support. A histogram with broad bins does not justify exact endpoints or a highly precise typical value. Use approximate language that matches the scale.
- Inferring causes from visible patterns. A graph can show a cluster, gap, or long tail; it cannot establish why that feature occurred.
- Treating a separated point as a mistake. A point apart from the rest is unusual in the display, but that alone does not show it is erroneous. Describe it and, if relevant, recommend checking the original record.
- Reading shape from a boxplot alone. A boxplot summarizes selected locations and spread but hides the detailed distribution within the quartiles. Use a graph that displays individual values or frequencies to describe shape.
Check Your Understanding
For each question, focus on what the graph supports and how to state it in context.
- A histogram of delivery times has a long tail toward larger values. What does this say about the distribution’s shape, and what does it not establish about the reason for the tail?
- A dotplot shows two clear groups of measurements with a gap between them. Why might one typical value fail to describe the distribution well?
- A boxplot shows a median of about 18 minutes and a middle-half interval from about 14 to 23 minutes. Write one contextual sentence reporting both location and variability.
- A graph’s horizontal axis runs from about 10 to 50 units, and its bins are 10 units wide. Why should a description avoid reporting a center to several decimal places?
- A single point is far from the main group in a dotplot. What can you say about the point, and what conclusion about its accuracy would require more evidence?