On this concept: Formula

Not sure which lesson to do first? Find your level and get a plan →

Math · Statistics & Probability · Grade 9-12

Type I and Type II Errors

Type I error (α) is rejecting H0 when it is actually true; Type II error (β) is failing to reject H0 when it is actually false.

α=P(reject H0∣H0 true), β=P(fail to reject H0∣H0 false)

Quick Answer

Type I error (α) is rejecting H0 when it is actually true; Type II error (β) is failing to reject H0 when it is actually false. Use these names whenever you ask what a wrong test decision would cost: a false alarm (a healthy patient flagged as sick, an innocent person convicted) or a miss (a real disease overlooked, a guilty person acquitted). The significance level α is the Type I error rate you choose; with the sample size fixed, lowering α raises β, and only more data lowers both.

Why This Matters

Every test decision can be wrong in two directions, and the two mistakes rarely cost the same. A false alarm on a cancer screening means extra tests and worry; a miss can let a disease grow untreated. Naming the errors lets you choose α by consequences rather than habit, and explains why a study with too few subjects can miss a real effect entirely.

Intuitive Explanation

A smoke alarm makes a decision every second, with H0: "no fire." It can be wrong in two ways. If it shrieks over burnt toast, it rejected a true H0: a Type I error, a false alarm. If it stays silent during a real fire, it failed to reject a false H0: a Type II error, a miss. When it is right, it is either quiet with no fire or sounding during a fire; the chance of that second correct outcome is called power.

Lay the four cases out as a grid, with the truth (H0 true or false) across the top and the decision (reject or not) down the side. The two errors sit in opposite corners. An error's name depends on the truth and the decision together: "reject" is a mistake only if H0 was true, and "fail to reject" is a mistake only if H0 was false.

The two error rates pull against each other. A more sensitive alarm catches more fires (smaller β) but goes off for toast more often (larger α). In a test, lowering α from 0.05 to 0.01 makes false alarms rarer but demands stronger evidence, so real effects are missed more often. The way to lower both at once is better information: a larger sample shrinks the standard error, so true and false nulls produce clearly different data.

A p-value is not an error rate. One study with p=0.03 has not made "a 3% error"; it either made an error or it didn't. α is the long-run rate of Type I errors when H0 is true, a property of the rule, not of one result.

Think of it like this. A referee's whistle faces the same two errors. Calling a foul that never happened is a Type I error; letting a real foul go is a Type II error. A strict referee misses fewer fouls but makes more false calls. Only a better view of the play, like video replay, cuts both, just as a larger sample does for a test.

Core idea. A Type I error rejects a true null (false positive); a Type II error fails to reject a false null (false negative).


When to Use

Use Type I and Type II language when a question asks what a wrong decision would mean, which mistake is worse, or how changing the significance level or the sample size affects the risks. Fix H0 first, then match the decision against the truth: rejecting a true H0 is Type I, keeping a false H0 is Type II. If the question asks how unusual one study's data are under H0, that is a p-value. If it asks how often a test catches a real effect, that is power, 1−β.

Tip. Ask: Am I classifying a wrong decision by comparing what the test concluded against what is actually true?

How to Recognize It

Write H0 in words first, then note what the test decided and what is actually true, because the name of an error depends on both.

  1. Is H0 actually true in the situation described, while the test rejected it?

    Yes is a Type I error, a false alarm: a fair coin declared biased, or an innocent defendant convicted when H0 is innocence. If H0 is false and the test kept it, that is a Type II error, a miss.

  2. Did the test reject an H0 that really is false?

    Yes is a correct decision, not an error, and its probability is the power, 1−β. Only the two mismatches between decision and truth get the names Type I and Type II.

  3. Does the question ask for a probability that was set before the test rather than computed from the sample?

    Yes, for a false alarm, means α, the Type I error rate the researcher chose. A probability computed from this sample's own result, measuring how surprising it is under H0, is the p-value.

  4. Does the question ask which error is worse in this setting?

    Yes means describe each error in context and what it would cost. Missing a dangerous disease (Type II, when H0 is no disease) is usually worse than a false alarm that leads to a second test, which argues for a larger α there.

Type I and Type II Errors compared with Power of a test, P-value, and Significance level α

All four rows connect to a hypothesis test's final decision. Ask whether you are naming an event or a probability: the wrong decisions (Type I and II errors), the chosen false-alarm rate (α), the chance of catching a real effect (power), or one sample's surprise (p-value).

Type I and Type II Errors

Meaning
Use this when you need to name or weigh the two ways a hypothesis-test decision can be wrong against the unknown truth.
Key test
Am I classifying a wrong decision by comparing what the test concluded against what is actually true?
Formula
α=P(reject H0∣H0 true), β=P(fail to reject H0∣H0 false)
Example
A test screens for a disease. H0: patient is healthy. The test flags a healthy patient as sick. Which error is this?

Power of a test

Meaning
The chance of correctly rejecting a false null — the good outcome, not the error.
Key test
Use when measuring the test's ability to detect a real effect.
Formula
Power =1−β
Example
A test that catches a true 10-point drug effect 90% of the time

P-value

Meaning
The rarity of one sample's data under the null, not a category of wrong decision.
Key test
Use when measuring how surprising a specific dataset is.
Formula
P(data∣H0)
Example
p =0.03 for this experiment

Significance level α

Meaning
The pre-set probability you allow for a Type I error; the error is the event, α is its rate.
Key test
Use when stating the chosen false-positive rate, not the event itself.
Formula
α=P(Type I)
Example
Setting α=0.05

Formula & Notation

α=P(reject H0∣H0 true), β=P(fail to reject H0∣H0 false)
α=P(reject H0∣H0 true); β=P(fail to reject H0∣Ha true); Power =1−β

How to read it: α=P(Type I error)=P(reject H0∣H0 true). β=P(Type II error)=P(fail to reject H0∣H0 false). Power =1−β.

Worked Examples

Example 1 — Disease screening

Easy

A test screens for a disease. H0: patient is healthy. The test flags a healthy patient as sick. Which error is this?
Show the steps and the answer
  1. Fix the truth: the patient is healthy, so H0 is true.

    An error's name depends on what is actually the case, so start there.

  2. Fix the decision: flagging the patient as sick means the test rejected H0.

    Declaring disease is acting against the default claim of health.

  3. Match them: rejecting a true H0 is a Type I error, a false positive.

    Its probability is α; at α=0.05, about 50 of every 1,000 healthy people screened would be flagged.

Answer: Type I error

Takeaway: Type I is a false alarm: the test rejects a null that was true.

Example 2 — Missed real disease

Standard

Same screening test, but now it tells a genuinely sick patient they are healthy. Is that also Type I?
Show the steps and the answer
  1. Fix the truth: the patient is sick, so H0 ('healthy') is false.

    This is what changed from the previous case; the decision alone never names the error.

  2. Fix the decision: 'healthy' means the test failed to reject H0.

    The test kept the default claim.

  3. Match them: failing to reject a false H0 is a Type II error, a false negative.

    If the test catches 90% of sick patients, then β=0.10: 1 sick patient in 10 is sent home.

Answer: No — this is a Type II error

Takeaway: Type II is a miss: something real was there and the test kept H0.

Example 3 — Find the mistake: lowering alpha cuts both errors

Application

A bottling line is tested with H0:μ=500 mL against Ha:μ<500, using n=25 bottles with σ=5 mL. The manager wants fewer false alarms and fewer missed underfills. Ana writes: "Lower α from 0.05 to 0.01; that cuts both errors." Suppose the true mean has drifted to 498 mL. Find the mistake and correct it.
Show the steps and the answer
  1. Cutoff at α=0.05: the standard error is 525=1, so reject when xˉ<500−1.645(1)=498.355.

    A one-sided test at 5% rejects the lowest 5% of sample means expected under H0.

  2. β when μ=498: P(xˉ≥498.355)=P(Z≥0.355)≈0.36.

    A Type II error happens when the drifted machine still produces a mean above the cutoff.

  3. Repeat at α=0.01: cutoff 500−2.326(1)=497.674, so β=P(Z≥−0.326)≈0.63.

    The stricter cutoff makes false alarms rarer but misses the real drift almost twice as often.

  4. Lower both with more data: with n=100 the standard error is 0.5, the cutoff is 500−2.326(0.5)=498.837, and β=P(Z≥1.674)≈0.05.

    A smaller standard error separates 498 from 500 clearly, so α and β can both be small.

Answer: Lowering α to 0.01 raises β from about 0.36 to about 0.63. To cut both errors, test more bottles: with n=100, α=0.01 and β≈0.05.

Takeaway: With the sample size fixed, α and β trade off; only more data lowers both.

Where You'll See It

  • Airport baggage screening

    With H0: 'this bag is safe', a Type I error means searching an innocent traveler's bag, and a Type II error means letting a dangerous item through. Because a miss is far worse, scanners are set to flag many harmless bags, accepting a high false-alarm rate to keep β tiny.

  • Email spam filters

    With H0: 'this email is legitimate', a Type I error sends a real message, perhaps a job offer, to the spam folder, and a Type II error lets junk into the inbox. Many filters lean toward Type II errors, because a lost job offer costs more than deleting a few junk emails.

  • Criminal trials

    With H0: 'the defendant is innocent', convicting an innocent person is a Type I error and acquitting a guilty one is a Type II error. The standard 'beyond reasonable doubt' sets a very small α, deliberately accepting more Type II errors to protect the innocent.

Common Mistakes

  1. The slip: Swapping Type I and Type II

    Instead: Type I rejects a TRUE null (false alarm), Type II misses a FALSE null (missed detection).

  2. The slip: Believing you can shrink both errors at once for fixed sample size

    Instead: lowering α raises β; only more data shrinks both.

  3. The slip: Calling a correct rejection an error

    Instead: rejecting a false null is power (1−β), the desired outcome, not a mistake.


Mini Practice

Try these on your own. Tap Reveal when you want to check.

  1. A smoke alarm tests H0: there is no fire. The alarm rings while you are only boiling water. Which error is this?

    Hint: Was H0 actually true? Did the alarm reject it?

  2. A court tests H0: the defendant is innocent. A guilty defendant is found not guilty. Which error is this?

    Hint: Was H0 actually false? Did the court reject it?

  3. A lab runs 200 tests at α=0.05, and in every one H0 is actually true. About how many Type I errors should it expect?

    Hint: α is the chance of rejecting a true null in each test.

  4. A test has β=0.25. In 80 studies where H0 is actually false, about how many will miss the effect, and what is the power?

    Hint: β is the miss rate when H0 is false. Power is what is left.

  5. A report says: "If the new tutoring program really works, our study will detect it 85% of the time." Is 0.85 a Type II error rate or the power?

    Hint: Is 0.85 the chance of catching a real effect or of missing it?

  6. H0: the parachute is safe. An inspector passes a parachute that is actually faulty. Sam says, "That is a Type I error, because it is the more dangerous one." Correct Sam.

    Hint: Decide whether H0 was true or false, then what the inspector did.

Want the full set?

50 practice questions for this concept — free to try, every one with a complete worked solution showing the why, not just the answer.

Frequently Asked Questions

Why can't we choose β in advance the way we choose α?

Because β\beta depends on the true value of the parameter, which you do not know. If a coin really lands heads 51% of the time, a test of 100 flips will almost always miss it; if it lands heads 80% of the time, the test will almost never miss. So β\beta is worked out for a specific alternative you care about, and that is how studies plan their sample size.

Is a false positive on a medical test the same as a Type I error?

It matches when H0H_0 is that the patient does not have the disease. Then a positive result for a healthy person rejects a true H0H_0, a Type I error, and a negative result for a sick person is a Type II error. Flip H0H_0 and the names flip too. That is why you state H0H_0 first and name the errors second.

Does a p-value of 0.03 mean there is a 3% chance I made a Type I error?

No. If you reject, either H0H_0 was true and you made a Type I error, or it was false and you did not; the p-value cannot tell you which. The rate α\alpha belongs to the procedure: when H0H_0 is true, a test at α=0.05\alpha=0.05 rejects it 5% of the time. The chance that one particular rejection is wrong also depends on how often the hypotheses you test are true.

Why is a Type I error usually treated as the more serious one?

By convention H0H_0 is the cautious default, such as a new drug working no better than the old one, and wrongly rejecting it leads to action: approving a useless drug, or changing a factory process for nothing. That is why α\alpha is fixed and kept small while β\beta is often allowed to be larger, around 0.2. In disease screening the priority reverses, and the miss is the error to avoid.

Learning Path

Type I and Type II Errors

You are here

Next →

Power of a Test
You need Hypothesis Testing and the P-Value first: a test decides by comparing p with α. This concept adds the four-outcome grid, the meaning of α as the false-alarm rate and β as the miss rate, and the trade-off between them. Next, the Power of a Test, 1−β, measures how often a test catches a real effect and shows how a larger sample raises it.

See Also