Statistics in the Media
Identify statistically informed arguments presented in traditional and digital media; discuss and critique methods, data representations and conclusions.
Read this first
A number in a headline was chosen by someone. Ask who was asked, and compared with what.
A bar chart can make a small difference look huge.
Words you need
- Claim
- Something someone says is true.
- Sample
- The group of people asked.
- Misleading
- Made to look different from the truth.
Worked example
e.g. A headline says nine out of ten people prefer Brand A. What should you ask?
- Ask how many people were asked. Ten? Or ten thousand?
- Ask who asked them. Was it Brand A itself?
- Ask what the other choice was.
- Prefer to what? Maybe the other option was much worse.
- Without those answers, the claim proves nothing.
A graph that does not start at zero makes small gaps look big.
Practice
The pictures fade as you go. By the last few you are on your own.
1Does this graph show a big difference?with a picture
AnswerNo. Only 4 apart out of about 90.
2A graph starts at 85 instead of 0. What happens?with a picture
AnswerThe gap looks much bigger than it is.
3A survey asked 8 people. Is that enough?with a hint
Need a hint?
How many people are there really?
AnswerNo. Far too few.
4Sales doubled. Is that impressive?with a hint
Need a hint?
What was the starting number?
AnswerNot if it went from 2 to 4.
5A dog food company says its food is best. What do you ask?on your own
AnswerWho did the test, and against what?
6Why should a bar chart start at zero?on your own
AnswerOtherwise the differences look bigger than they are.
7Name one question to ask about any statistic.on your own
AnswerHow many people were asked?
Think about it
1
Why does the size of the sample matter?
One good answerA few people cannot speak for everyone.
2
Why ask who paid for a study?
One good answerThey may have wanted a particular answer.
Common mistakes
Believing a number because it is printed.
Ask where it came from first.
Reading bar heights without checking the scale.
A cut-off scale exaggerates the gaps.