Skip to lesson

08 / 12 · Learn from a sample

Why does a small sample jump around?

Understand why larger random samples usually give steadier proportions, not guaranteed exact answers.

Builds on Who actually answered?

Go to practice ↓

A question to keep in mind

Does 7 heads in 10 fair tosses mean the coin's true chance is 70%?

A sample proportion is a result from this run. Change the run and the result can change, even when the underlying chance stays the same. This is sampling variation, not necessarily a calculation error.

With independent draws from the same process, larger samples usually make proportions less variable. They do not guarantee an exact answer, and do not repair a biased way of choosing people.

Work through an example

RunOut of 10Out of 100
13 (30%)45 (45%)
27 (70%)55 (55%)
34 (40%)48 (48%)
46 (60%)52 (52%)
Constructed examples of possible sample results, not a promise for every run.
  1. Seven heads out of ten gives a sample result of 70%.
  2. The fair-coin assumption still says the chance of each independent toss is 50%.
  3. The gap is 20 percentage points. A short run can differ from the underlying chance.

Your turn

0 / 3
Practice 1Not checked

Over many repeated runs of independent fair-coin tosses, which sample proportion usually varies less?

Your answer
Read solution · does not award completion

The proportion from 100 tosses.

This is a statement about repeated sampling, not a claim that every 100-toss run is closer than every 10-toss run.

Practice 2Not checked

One run has 4 heads out of 10; another has 52 out of 100. How far is each percentage from 50%, in percentage points?

Your answer
Read solution · does not award completion
  • Gap for 4 out of 10: 10 points
  • Gap for 52 out of 100: 2 points

The gaps are 10 and 2 percentage points. In this particular example the larger sample is closer; no single example proves the general pattern.

Practice 3Not checked

A website surveys only its most active readers. Would collecting ten times as many of those readers automatically represent all readers?

Your answer
Read solution · does not award completion

No. The same selection can still miss less-active readers.

A larger sample can reduce random variation while leaving selection bias in place.

Bring it back to your own work

Explain to a friend why 'more responses' and 'a better survey' are not always the same thing. Use one concrete example.

Saved only on this browser0 / 6000