Type I vs Type II Errors: A Simple Way to Remember the Difference

The two ways a test can go wrong

Every hypothesis test ends with one of two decisions: reject the null hypothesis, or fail to reject it. Since you never know the truth, each decision carries a way of being wrong.

• A Type I error is rejecting a null hypothesis that was actually true. You announce an effect that is not there. A false positive.
• A Type II error is failing to reject a null hypothesis that was actually false. You miss a real effect. A false negative.

The memory trick that actually works

Forget the numbers for a moment and think about a fire alarm.

The alarm goes off when there is no fire — that is a false alarm, a Type I error. The alarm stays silent while the building burns — that is a miss, a Type II error.

Now the ordering makes sense: Type I comes first because it is the alarm that fires first. One is a noisy mistake, the other a silent one — and the silent mistakes are usually the dangerous ones.

Alpha, beta and power

Each error has a probability attached, and one more term completes the picture:

• α (alpha) is the probability of a Type I error. You choose it — usually 0.05. Setting α = 0.05 says: I accept a 5% chance of crying wolf.
• β (beta) is the probability of a Type II error. You do not choose it directly; it follows from your sample size, your significance level, and how big the real effect is.
• Power = 1 − β is the probability of correctly detecting a real effect. Convention treats 80% as the minimum acceptable.

The trade-off you cannot escape

Here is the part students find genuinely counterintuitive: reducing one error rate increases the other, if everything else stays fixed.

Tighten α from 0.05 to 0.01 and you demand stronger evidence before declaring an effect. Fewer false alarms — but you will now miss real effects you would previously have caught. β rises and power falls.

Loosen α and the reverse happens. The only way to improve both at once is to collect more data. That is the real reason sample size matters, and why a power calculation belongs before a study, not after it.

Which error is worse? It depends

Statistics cannot answer this. It is a judgement about consequences.

In a medical screening test, a Type II error means telling a sick patient they are healthy — potentially fatal. A false positive means an anxious week and a second test. Here you accept more false alarms to avoid misses.

In a criminal trial, where the null hypothesis is innocence, a Type I error convicts an innocent person. Most legal systems set α extremely low — “beyond reasonable doubt” — and accept that guilty people will sometimes go free.

In quality control, a Type I error halts a production line unnecessarily; a Type II error ships defective goods. The right balance depends entirely on the relative cost.

How to raise power

If you want a better chance of detecting a real effect, you have four levers — and only three are under your control:

1. Increase the sample size. The most reliable option, and the only one with no downside beyond cost.
2. Increase α. Works, but buys power with more false positives. Rarely a good trade.
3. Reduce variability. Better measurement, tighter controls, or a paired design instead of two independent groups.
4. Hope the effect is large. Not a lever at all — effect size is a fact about the world, not a choice.

The mistake nobody warns you about

Running a study, getting a non-significant result, and concluding “there is no effect” is a Type II error waiting to happen. If the study was underpowered, a null result was always the likely outcome — effect or no effect.

Before believing any negative finding, ask what power the study had. A study with 30% power that finds nothing has told you almost nothing.

Summary table

Learn this table properly and a whole class of exam questions becomes mechanical. The rest is judgement about which mistake you can least afford.