What is the difference between a Type I and a Type II error in simple terms?

A Type I error is a false alarm — you reject the null hypothesis when it is actually true. A Type II error is a missed detection — you fail to reject the null when it is actually false. The probability of a Type I error is ; the probability of a Type II error is .

How do you increase the power of a hypothesis test?

Increase the sample size, increase , study a larger true effect, or reduce variability in the data. The safest choice that does not raise the false-alarm rate is increasing the sample size, since it lowers the standard error and reduces .

No. is computed assuming the null hypothesis is true, while is computed assuming a specific alternative is true. They describe two different scenarios, so they are not complements. Power and do add to 1 because Power = 1 -.

Which type of error is worse?

Neither is universally worse — it depends on the real-world consequences. If a false positive is costly (like approving a harmful drug), the Type I error is more serious. If missing a real effect is costly (like failing to detect a disease), the Type II error is worse. On the exam, justify your answer using the specific context.

AP-STATS-6.7

U6.7 Type I, Type II, and Power

Master Type I and Type II errors and statistical power in AP Statistics: definitions, probabilities, trade-offs, and how the exam tests each concept.

What you'll do in this lesson

A voice-first session with the Crimsora tutor on U6.7 Type I, Type II, and Power, then targeted practice and FRQs — with the tutor adapting to where you get stuck.

Start U6.7 Type I, Type II, and Power free

What this lesson covers

Every significance test forces a decision under uncertainty, and sometimes that decision is wrong. In this lesson you will learn the two ways a hypothesis test can go wrong — a Type I error and a Type II error — and how to describe each in the context of a real problem. You will also meet power, the probability that a test correctly detects a true effect. These ideas connect directly to significance level

\alpha

, sample size, and effect size, and the AP exam loves to ask you to interpret them in plain language. By the end you will be able to name the error, state its consequence, and reason about what raises or lowers power.

The Two Errors: Type I and Type II

A hypothesis test starts with a null hypothesis

H_0

and an alternative

H_a

. Because we base the decision on sample data, we can reach the wrong conclusion in two distinct ways.

A Type I error happens when we reject

H_0

even though

H_0

is actually true. We announce an effect that does not exist — a false alarm. The probability of a Type I error equals the significance level,

P(\text{Type I}) = \alpha

.

A Type II error happens when we fail to reject

H_0

even though

H_0

is actually false. We miss a real effect. Its probability is denoted

\beta

.

The key is that the truth about

H_0

is fixed but unknown; the error depends on which decision we make. Use this table to keep them straight.

Truth	We reject $H_0$	We fail to reject $H_0$
$H_0$ true	Type I error ( $\alpha$ )	Correct
$H_0$ false	Correct (power)	Type II error ( $\beta$ )

A common misconception is thinking a Type I error means "the null is false." It does not — a Type I error only occurs when

H_0

is genuinely true. Always anchor your description to the assumed truth of

H_0

Describing Errors in Context

The AP exam almost never asks "what is a Type I error?" in the abstract. Instead it gives a scenario and asks you to describe the error and its consequence in context. To earn full credit you must do three things: identify what

H_0

claims, state the wrong decision, and explain the real-world cost.

Suppose a company tests whether a new drug is more effective than a placebo. Let

H_0

: the drug is no better than placebo, and

H_a

: the drug is better.

A Type I error means concluding the drug works when it truly does not. The consequence: the company markets an ineffective drug, wasting money and possibly exposing patients to side effects for no benefit.

A Type II error means concluding there is not enough evidence the drug works when it actually does. The consequence: a genuinely helpful drug is abandoned, and patients miss out on real benefit.

Notice how the consequences differ. Deciding which error is "worse" depends on context, and the exam sometimes asks you to argue this. There is no universally worse error — it depends on the costs involved. Write your answer using the specific nouns from the problem, not generic phrases like "reject the null," which will not earn context points.

Power and What Affects It

Power is the probability that a test correctly rejects a false null hypothesis. In symbols,

\text{Power} = 1 - \beta

, where

\beta

is the probability of a Type II error. High power means the test is good at detecting a real effect of a given size.

Four factors influence power, and you should know the direction of each.

Change	Effect on power
Increase sample size $n$	Power increases
Increase $\alpha$ (e.g. 0.05 to 0.10)	Power increases
Larger true effect (farther from $H_0$ )	Power increases
Less variability in the population	Power increases

The most reliable way to raise power without accepting more false alarms is to increase the sample size. Increasing

\alpha

also raises power, but at a cost we discuss next. Effect size and variability are usually properties of the situation rather than choices you control.

A frequent exam trap: students say power "is the probability the null is false." No — power is a conditional probability computed assuming a specific true alternative value. Power always references a particular effect size, because a test detects a large effect more easily than a tiny one.

The Trade-Off Between the Errors

You cannot make both error probabilities as small as you like at the same time, holding sample size fixed. Lowering

\alpha

makes it harder to reject

H_0

, which reduces the chance of a Type I error but increases

\beta

and therefore lowers power. Raising

\alpha

does the opposite.

Think of

\alpha

as a threshold. A strict threshold (small

\alpha

) demands strong evidence before rejecting, so false alarms are rare but real effects are missed more often. A lenient threshold (large

\alpha

) catches more real effects but also triggers more false alarms.

The relationships to memorize:

P(\text{Type I}) = \alpha, \qquad P(\text{Type II}) = \beta, \qquad \text{Power} = 1 - \beta.

The only way to reduce both

\alpha

and

\beta

simultaneously is to collect more data. That is why increasing

n

is the go-to answer when the exam asks how to improve a study without inflating the false-alarm rate.

A common misconception is treating

\alpha

and

\beta

as if they add to 1. They do not.

\alpha

is computed assuming

H_0

is true, while

\beta

is computed assuming a particular alternative is true — two different scenarios entirely.

How the Exam Frames These Questions

On the AP exam, Type I/II and power questions appear in both multiple-choice and free-response formats. Multiple-choice items often ask you to match a described mistake to the correct error type, or to predict how changing

n

\alpha

affects power. Free-response items typically embed the concept inside a larger inference problem and ask for a contextual description plus a consequence.

A reliable strategy is to first write the hypotheses clearly, then translate. "Reject a true

H_0

" equals Type I; "fail to reject a false

H_0

" equals Type II. Then attach the real-world outcome using the scenario's own language.

When asked about power, name the specific alternative value in play, because power is defined relative to a true parameter value. If a question asks how to increase power while keeping

\alpha

fixed, the expected answer is increasing the sample size (or, if offered, reducing variability or studying a larger effect).

Beware questions that ask which error is more serious. There is no automatic answer; you must justify your choice by comparing the concrete consequences described. Graders reward reasoning tied to the context, so avoid vague statements and always reference the actual stakes of the problem.

Key terms

Type I error.: Rejecting $H_0$ when $H_0$ is actually true — a false positive. Its probability equals the significance level $\alpha$ .
Type II error.: Failing to reject $H_0$ when $H_0$ is actually false — a missed detection. Its probability is denoted $\beta$ .
Significance level ( $\alpha$ ).: The threshold probability for rejecting $H_0$ , chosen before the test; also the probability of committing a Type I error when $H_0$ is true.
Power.: The probability a test correctly rejects a false $H_0$ , equal to $1-\beta$ , computed relative to a specific true alternative value.
Beta ( $\beta$ ).: The probability of a Type II error, computed assuming a particular alternative parameter value is true.
Effect size.: How far the true parameter lies from the null value; larger effect sizes make a false null easier to detect, raising power.

Worked example

A quality-control engineer tests whether more than 5% of manufactured chips are defective. She sets

H_0: p = 0.05

versus

H_a: p > 0.05

using

\alpha = 0.05

. Describe a Type I and a Type II error in context, state a consequence of each, and explain one way to increase the power of the test.

First identify what

H_0

claims: the defect rate is exactly 5% (not elevated). The alternative claims the rate exceeds 5%.

A Type I error means rejecting

H_0

when it is true — concluding the defect rate is above 5% when it really is 5%. Consequence: the engineer needlessly halts production or scraps a good batch, wasting time and money.

A Type II error means failing to reject

H_0

when it is false — concluding there is not enough evidence of an elevated defect rate when the rate truly is above 5%. Consequence: defective chips ship to customers, damaging reliability and reputation.

Power is the probability of correctly detecting a truly elevated defect rate,

1 - \beta

. To increase power while keeping

\alpha

at 0.05, the engineer can increase the sample size — inspecting more chips. A larger

n

reduces the standard error of

\hat{p}

, making it easier to distinguish a real increase from ordinary sampling variation, so

\beta

falls and power rises.

Note we described each error using the concrete nouns of the scenario (defect rate, batches, customers), which is what earns full credit.

Practice questions

A researcher tests

H_0: \mu = 100

against

H_a: \mu \neq 100

\alpha = 0.05

. Holding everything else constant, which change would increase the power of the test?

Decreasing the sample size
Decreasing $\alpha$ to 0.01
Increasing the sample size
Increasing the population standard deviation

Answer: Increasing the sample size

Power rises when the test can more easily distinguish a true effect from random variation. A larger sample size reduces the standard error, increasing power. Decreasing

n

or increasing variability lowers power, and decreasing

\alpha

makes rejection harder, which also lowers power.

A court analogy treats

H_0

as 'the defendant is innocent.' In this framing, describe what a Type I error and a Type II error represent, and explain why society might choose a small

\alpha

Answer: A Type I error is convicting an innocent defendant (rejecting a true $H_0$ ). A Type II error is acquitting a guilty defendant (failing to reject a false $H_0$ ). Society often chooses a small $\alpha$ to make wrongful conviction rare, accepting that this raises the chance of letting some guilty defendants go free.

The null is 'innocent,' so rejecting it means convicting. Rejecting a true null (convicting an innocent person) is the Type I error; failing to reject a false null (acquitting a guilty person) is the Type II error. Lowering

\alpha

demands stronger evidence to convict, reducing wrongful convictions but increasing wrongful acquittals — a direct illustration of the trade-off between the two error types.

A test currently has

\beta = 0.30

. What is its power, and what does that number mean in context of detecting a true effect?

Answer: The power is $1 - 0.30 = 0.70$ , meaning there is a 70% probability the test will correctly reject the null hypothesis when the specified true effect actually exists.

Power equals

1 - \beta

. Here

1 - 0.30 = 0.70

. This value is conditional: it assumes a particular true alternative value. A power of 0.70 means that if the effect really is that size, the test detects it 70% of the time and misses it 30% of the time (the Type II error rate).

FAQ

What is the difference between a Type I and a Type II error in simple terms?: A Type I error is a false alarm — you reject the null hypothesis when it is actually true. A Type II error is a missed detection — you fail to reject the null when it is actually false. The probability of a Type I error is $\alpha$ ; the probability of a Type II error is $\beta$ .
How do you increase the power of a hypothesis test?: Increase the sample size, increase $\alpha$ , study a larger true effect, or reduce variability in the data. The safest choice that does not raise the false-alarm rate is increasing the sample size, since it lowers the standard error and reduces $\beta$ .
Do $\alpha$ and $\beta$ add up to 1?: No. $\alpha$ is computed assuming the null hypothesis is true, while $\beta$ is computed assuming a specific alternative is true. They describe two different scenarios, so they are not complements. Power and $\beta$ do add to 1 because $\text{Power} = 1 - \beta$ .
Which type of error is worse?: Neither is universally worse — it depends on the real-world consequences. If a false positive is costly (like approving a harmful drug), the Type I error is more serious. If missing a real effect is costly (like failing to detect a disease), the Type II error is worse. On the exam, justify your answer using the specific context.

Learn this with a teacher, not a page

The Crimsora tutor teaches U6.7 Type I, Type II, and Power live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.

Try a live session free