MATH120 Guide: Choose Descriptive Summaries for a Distribution

Start with the descriptive job

Descriptive statistics organize observed data. Decide whether the goal is to show composition, compare groups, locate a typical value, quantify variability, reveal distribution shape, or flag unusual cases. These jobs require different evidence. A compact table may answer a category-composition question better than a complicated graph, while one average cannot answer a question about consistency.

Write the interpretation target before selecting the display: “I need to compare the distribution of ___ across ___” or “I need to describe how ___ varies among the observed ___.” This prevents software defaults from dictating the story. The eventual conclusion should use the same variable, group, unit, and timeframe.

Summarize categorical data with counts and defensible denominators

A frequency table reports how many observations fall into each category. Relative frequencies or percentages describe the share of a defined total. Always state the denominator, especially when missing values, multiple selections, eligibility rules, or subgroup filters change it. A bar chart preserves distinct categories and allows comparison without implying a continuous numeric scale.

Order categories logically or by frequency when that supports reading, but do not imply magnitude between nominal labels. For ordinal categories, preserve the natural order while remembering that gaps between levels are not necessarily equal. When comparing groups of different sizes, counts alone may mislead; proportions can improve comparability if their denominators and sample contexts are clear.

Use a display that reveals quantitative shape

A histogram groups quantitative values into intervals and shows distribution shape, while a dotplot or stem-and-leaf display can preserve more individual detail for smaller datasets. A boxplot compresses center, quartiles, spread, and flagged extremes and is useful for group comparisons, but it hides modes, gaps, and detailed shape. Choose according to the question and amount of data.

Inspect axes, bin choices, scale, truncation, units, and grouping. Different bin widths can make the same values look smoother, noisier, unimodal, or multi-peaked. A graphical pattern is descriptive evidence, not proof of a cause. Record which feature persists across reasonable display choices.

Interpret mean and median as different centers

The mean is the balancing value determined by every magnitude. It moves toward extreme observations and long tails. The median is the ordered halfway point and is resistant to how far an extreme lies from the rest. Neither is universally superior. The choice depends on the question, distribution, and meaning of typical.

In a roughly balanced unimodal distribution without influential extremes, mean and median may tell a similar story. In a strongly right-skewed distribution, the mean often exceeds the median because high values pull the balance point. Reporting both can reveal that asymmetry. Explain the mechanism instead of applying a memorized rule without inspecting the data.

Pair center with spread

A center does not say whether observations cluster tightly or vary widely. Range uses only the minimum and maximum and is sensitive to extremes. The interquartile range describes the width of the middle half of ordered values and pairs naturally with the median. Standard deviation describes a typical distance-like scale around the mean, but its interpretation should be tied to the distribution and units.

Compare spread only when groups use compatible variables, units, and contexts. A larger standard deviation may reflect genuine heterogeneity, mixed subgroups, measurement differences, or extreme cases. It does not automatically mean poor quality. State what varies and why that variation matters to the question.

Describe shape before compressing it

Look for symmetry or skew, one or several peaks, clusters, gaps, boundaries, and unusual observations. These features determine whether a chosen center and spread summarize the distribution faithfully. A bimodal pattern may indicate meaningful subgroups that one average conceals. A hard lower bound can create skew even without errors.

Use careful language: “The observed distribution is right-skewed” describes recorded values; it does not establish why. Investigate whether unusual points are genuine, entry errors, different populations, or rare but relevant cases. Do not remove them solely to make the summary easier. Report sensitivity when their inclusion materially changes the interpretation.

Cross the Calculation-to-Interpretation Bridge

A calculation statement names the value and method: “The observed median was ___ units.” An interpretation statement explains what position or comparison that value represents for the observed group. A limitation statement identifies what the summary omits or cannot establish. Keeping these sentences separate exposes unsupported leaps.

For a proportion, name the numerator category and denominator group. For a mean, include the variable and unit. For a median, explain the halfway position rather than claiming every typical observation equals it. For spread, state whether values are tightly or widely distributed relative to a relevant scale without inventing a threshold.

Fictional example: one extreme changes the center story

Imagine fictional repair costs where most recorded jobs are modest and one specialized repair is extremely expensive. The high value can raise the mean substantially while leaving the median position relatively stable. That does not make the high cost wrong or irrelevant. It tells the learner that the distribution is right-skewed and that “typical” needs clarification.

A responsible description might present median and interquartile range for the middle pattern, mention the extreme case, and optionally report the mean when the total-cost balancing perspective matters. The example contains no assignment values or completed analysis; the learner must make the choice for their own question and data.

Compare groups with common definitions and parallel evidence

Before comparing centers or proportions, confirm that groups use the same variable definition, unit, timeframe, eligibility rules, and missing-data treatment. Then examine distributions, not just a difference in averages. One group may have the same median but much greater spread, or a higher mean caused by a small tail. Sample sizes and source quality also affect what the observed comparison can support.

Use parallel wording for each group and describe magnitude without turning a descriptive difference into a causal effect. If groups arise from observational data, lurking variables and selection may explain part of the pattern. A comparison is a starting point for interpretation, not automatic proof of a mechanism.

Run the Summary-and-Shape Selector

Check five connections: variable meaning to valid operations, question to descriptive job, job to display, display to center and spread, and summary to conclusion. If any link breaks, revise the representation rather than decorating it. Confirm units, category denominators, axis scales, missing values, and the influence of unusual observations.

Read the conclusion without the graph. Does it still name the observed group, variable, pattern, variation, and limitation? Then read the graph without the conclusion. Are labels and scales sufficient for an independent reader? Agreement between the two is a practical verification that calculation and interpretation remain connected.

Repeat the check after changing one defensible display choice, such as a nearby histogram interval width or the inclusion of a transparently identified unusual observation. If the main story changes sharply, report that sensitivity instead of selecting the version that looks most persuasive. Verify group comparisons with identical scales and definitions, and make sure every percentage names its denominator. These steps separate a robust descriptive pattern from an artifact of presentation choices. Record the reason for the final display so another reader can reproduce the interpretive path and assess whether the conclusion remains supported.

Use summary guidance without requesting a finished analysis

Domyclass can explain why a display or statistic fits a data type, help interpret learner-owned output, or suggest checks for skew and unusual observations. It will not analyze a current graded dataset, produce submission-ready graphs or conclusions, or choose live assessment answers. The learner selects the representation, calculates the summaries, evaluates limitations, and writes the final interpretation.

Related MATH120 resources

• MATH120 Introduction to Statistics Help for AMU Students
• When is the median more informative than the mean?
• MATH120 Guide: Turn a Statistical Question into a Data and Variable Map
• MATH120 Questions and Answers