The University of Melbourne · FACULTY OF STATISTICS

CMCE10002 Chap.5 Why Summary Statistics Alone Can Mislead

- one subject, every graph, every model, every mark
5 Chapters3-page Bible
Our own words - no uploaded lecturer files
Updated for this semester
Chapter 5 of 13 · CMCE10002

Why Summary Statistics Alone Can Mislead

The summary table agreed and the data did not

Week 3 does not open with a chart function.

It opens by asking why visualisation should come first, and answers by putting a block of separate datasets beside one another and showing that their summary statistics are effectively identical while the pictures are nothing alike.

What a summary statistic is allowed to say

An average locates a distribution and a standard deviation says how spread out it is.

Neither says anything about shape: one cluster or three, a smooth rise or a bend, a handful of extreme rows carrying the result.

Two of those decide whether a recommendation is sound, and none of them is visible in the table.

The examinable version

The habit this builds is a sentence rather than a picture: say what a table of averages supports, name the specific thing it hides, and give the step that would settle the question.

Where this returns later in the subject

Grouped summaries inherit the same weakness in concentrated form: the smaller the group, the more one unusual row moves its average, which is why a count belongs beside every group average.

Causal work supplies the other half of the argument, the comparison group that lets a difference between two averages carry a causal reading. Read those two later chapters as the answers to the two failure modes named here, one about hidden spread and one about a rival explanation.

In this chapter

What this chapter covers

  • 01

    Why the week 3 lecture puts visualisation before analysis

  • 02

    Datasets that share a mean and a standard deviation while differing completely

  • 03

    What an average and a spread can and cannot establish

  • 04

    Reading a table of group averages without overreaching

  • 05

    Distinguishing a spurious pattern from a hidden spread

Worked example · free

Reading a table of group averages honestly

Q [3 marks]. AskSia assigns three practice points to this independent exercise; they are not a University marking scheme. Products are grouped into six average-rating bands and each band shows higher average weekly sales than the one below it. A manager asks whether the business should invest in raising ratings.
  • 1State what the table does support.
  • 1State the two separate reasons it does not answer the question.
  • 1Give the next analytical step.
The table supports a monotone descriptive pattern: products with higher average ratings also record higher average weekly sales, and the gap between consecutive bands is not constant. It does not support the investment decision for two separate reasons. Products in different bands differ in other ways as well, including price, category, brand strength and promotion, any of which could produce the pattern; and band averages hide the spread within each band, so a band average can be produced by a few very large sellers. The next step is to compare products that are alike in those other respects, or to follow products whose ratings actually changed and see what their sales did afterwards.
Sia tip — Name which failure you mean. A pattern driven by something else needs a better comparison; a summary hiding its spread needs the underlying distribution. They are different repairs.
Glossary

Key terms

Summary statistic
A single number standing in for a whole distribution, such as a mean or a standard deviation.
Standard deviation
A measure of how widely the values in a distribution are spread around their centre.
Distribution shape
How observations are arranged across the range, which two identical summary numbers can hide completely.
Spurious pattern
An association produced by something other than the relationship being claimed.
FAQ

Why Summary Statistics Alone Can Mislead FAQ

Can two datasets have the same average and still look completely different?

Yes, and the week 3 lecture opens with exactly that demonstration. Several datasets can be constructed to share a mean and a standard deviation on both axes while their scatterplots have nothing in common, which is why the subject plots before it summarises.

What should a short answer about a table of averages contain?

Three clauses: what the table supports, the specific thing it hides or the specific alternative explanation, and the next analytical step. Stopping after the first reads as description and stopping after the second reads as scepticism.

Why does the subject teach visualisation before analysis?

Because a chart is a diagnostic step rather than a presentational one. Plotting first tells you which summary statistics are safe to quote, and occasionally tells you that none of them are, which is information you cannot recover from the summary table itself.

Study strategy

Exam move

Whenever a table of averages appears in practice material, write the three-clause response before reading on. Supported, hidden, next step. It is the shape most interpretation questions in this subject reward.

Practise on any table that reports one number per group: write down what you would need to see to be confident it is not a handful of rows driving the result, then say which column of the table would tell you and which one is missing.

Working through Why Summary Statistics Alone Can Mislead in CMCE10002? Sia is AskSia’s AI Statistics tutor — ask any CMCE10002 Why Summary Statistics Alone Can Mislead question and get a clear, step-by-step explanation grounded in how CMCE10002 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

A+Everything unlocked
Unlocks this Bible + all 74 of your The University of Melbourne subjects - and 1,000+ Bibles across every Australian university.
Sia - your CMCE10002 tutor, unlimited, worked the way the exam marks it
The full 3-page Bible + practice bank with worked solutions
Chrome extension - sync your LMS so Sia knows your deadlines
Bilingual EN / Chinese on every Bible and every Sia answer
$0.99 Trial
30-day money-back · cancel in one tap · how it works
Unlock the full CMCE10002 Bible + 74 The University of Melbourne subjects
$0.99 Trial