Topics Numbers and logic

How can statistics mislead you?

Statistics mislead you mostly without lying. A number can be perfectly accurate and still leave you with the wrong picture, because of what was measured, who was counted, what was left out, or how the result was framed. An average hides a spread, a percentage hides a starting size, and a chart can make a small change look dramatic just by where the axis begins.

What makes this interesting is that the tricks are old and ordinary. Correlation gets mistaken for cause. A sample of people who chose to answer a survey gets treated as everyone. A rare event sounds alarming as a relative change and trivial as an absolute one. Even careful people fall for these, because a tidy number feels like a fact, and we rarely ask where it came from.

An episode would walk through the main traps one at a time, with concrete examples, and end with a short set of questions you can ask of any statistic you meet. It is made by bre, whose hosts are AI and can be wrong, so treat it as a starting point and check anything that matters. You can also press Talk and ask about a number you saw.

What a bre episode would cover

An outline of the episode bre would make for this question. Every episode is written fresh when you ask, so yours will differ.

  1. Accurate numbers, wrong pictureWhy a statistic can be true and still mislead, and how choices about what to measure shape the story before any math happens.
  2. The trouble with averagesHow a mean can be dragged by a few extreme values, why the median and mode sometimes say more, and what an average hides about spread.
  3. Who got counted?Sampling and selection bias: surveys where only some people answer, and groups that are unrepresentative of the whole.
  4. Correlation is not causeTwo things moving together does not mean one drives the other. Hidden third factors, coincidence and reverse direction all play a part.
  5. Relative versus absolute changeWhy doubling a tiny risk is still tiny, and how percentages without a starting size can sound far bigger or smaller than they are.
  6. Charts that bend the truthTruncated axes, cherry-picked time windows and mismatched scales, and how they change what the same data seems to say.
  7. Questions to ask of any numberA short checklist: compared to what, out of how many, who was left out, and who benefits from this framing.

How the episode might open

A sample exchange between two of bre’s AI hosts, bre and Arlo. Both are AI; this is written by AI, as every bre episode is.

  1. breAI host

    Here's a sentence that sounds harmless: the average person in this room is a millionaire. Then one billionaire walks in, and suddenly nobody else's bank account has changed, but the number has.

  2. ArloAI host

    Average's doing the lying there.

  3. breAI host

    Kind of. The math is correct. It just answers a question nobody was asking, which was how much money is in the room, divided by heads.

  4. ArloAI host

    So use the median. Middle person. Done.

  5. breAI host

    Often, yes. But even the median can hide things, like whether everyone clusters together or the room splits into two very different groups. That's where this gets slippery.

  6. ArloAI host

    So no number's safe.

  7. breAI host

    Not quite. A number is safe when you know what it counted and what it skipped. Okay, here's the part nobody tells you: most misleading statistics aren't wrong, they're just missing context.

  8. ArloAI host

    Who counted that? That's my first question every time.

Questions people also ask

Can a statistic be true and still misleading?
Yes, and that is the most common case. The calculation can be correct while the choice of measure, sample, comparison or time window gives a skewed impression. Misleading usually means missing context, not false arithmetic.
What is the difference between correlation and causation?
Correlation means two things tend to change together. Causation means one actually produces the other. A link can come from coincidence, a hidden third factor, or the cause running the other way, so a correlation alone does not settle why.
Why do percentages sound scarier than they are?
A percentage change is relative to a starting size. If something very rare doubles, the percentage sounds huge, but the absolute change may be tiny. Knowing the actual counts, such as how many out of how many, gives a fairer sense of scale.
How can I spot a misleading statistic?
Ask what was measured, who was included, how big the sample was, what it is being compared to, and what is left out. Check whether the chart's axis is truncated and whether the source has a reason to favor one reading.

Related topics

More: all 300 topics, numbers and logic, or the longer reads on /learn.

bre’s hosts are AI, and every episode is generated, so they can be wrong: check anything that matters. This page outlines what an episode would cover. It is for interest and learning, not medical, financial or legal advice.