Unbiased Estimator in AP Statistics

In AP Statistics, an unbiased estimator is a statistic whose sampling distribution has a mean equal to the population parameter it estimates, so across many random samples it neither overestimates nor underestimates on average (for example, x̄ for μ and p̂ for p).

What is Unbiased Estimator?

An unbiased estimator is a statistic that is right on average. Picture taking every possible random sample of the same size, calculating the statistic each time, and plotting all those values. That plot is the sampling distribution. If its center (its mean) lands exactly on the true population parameter, the estimator is unbiased.

Any single estimate can still miss, sometimes by a lot. Unbiased doesn't mean "always correct." It means the misses are not systematically in one direction. Overestimates and underestimates balance out. The classic AP examples are the sample mean x̄ (unbiased for μ), the sample proportion p̂ (unbiased for p under random sampling), and the sample variance s² (unbiased for σ², but only because it divides by n − 1 instead of n).

Why Unbiased Estimator matters in AP® Statistics

This term lives in Topic 3.1 Estimators, where you judge whether a statistic does a good job of estimating a parameter. Bias is one of the two big criteria. The other is variability, meaning how spread out the sampling distribution is. Unbiasedness is also the quiet assumption behind confidence intervals and significance tests. Those procedures center on x̄ or p̂ because those statistics aim at the true value. If your estimator were biased, the interval would be built around the wrong spot, and no margin of error could fix that.

How Unbiased Estimator connects across the course

Sampling Distribution (Topic 3.1)

Bias is a property of the sampling distribution, not of any one sample. To check whether an estimator is unbiased, you look at where the whole distribution of possible estimates is centered. Think of it as asking "where does this statistic aim?" rather than "did this one shot hit?"

Biased Estimator (Topic 3.1)

A biased estimator's sampling distribution is centered off the parameter. A good example is dividing the sum of squared deviations by n instead of n − 1, which tends to underestimate σ². The n − 1 in s² is the fix that removes that bias. That's why your calculator uses it.

Point Estimate (Topic 3.1)

A point estimate is the single number you actually compute from your one sample, like x̄ = 72.4. Unbiasedness describes the method that produced that number, not the number itself. A good method can still hand you an unlucky estimate.

Is Unbiased Estimator on the AP® Statistics exam?

Expect this concept mostly in multiple choice, plus explanation-style reasoning on free response. Common tasks:

• Explain why a statistic is unbiased. A typical stem says a researcher claims x̄ is unbiased for μ and asks you to pick the correct reason. The answer is that the mean of the sampling distribution of x̄ equals μ.
• Compare s² to the divide-by-n version. You need to know that dividing by n − 1 makes s² unbiased for σ², while dividing by n gives a biased estimator that runs too small on average.
• Identify conditions. One question type asks when p̂ is unbiased for p when sampling without replacement from a finite population. Random sampling is what makes p̂ unbiased. The 10% condition matters for the standard deviation formula, not for bias.
• Check a constructed estimator. You might see something like T = x̄ + c(s² − σ²). Since s² is unbiased, the expected value of (s² − σ²) is 0, so T is unbiased for μ no matter what c is.
• Read simulated sampling distributions. You may get dotplots or histograms of estimates from several estimators and have to pick the one centered at the parameter, then weigh bias against spread.

No released FRQ has used the phrase verbatim, but free-response questions often ask you to evaluate an estimator, and "centered at the parameter" is the language that earns credit.

Unbiased Estimator vs Low variability (precision)

Unbiased means the sampling distribution is centered on the parameter. Low variability means the sampling distribution is tightly clustered. These are separate ideas. Picture a dartboard. Unbiased throws average out to the bullseye but can be scattered everywhere. Low-variability throws land close together but might cluster off to one side. The best estimator does both. Bigger samples reduce variability, but they do not remove bias.

Key things to remember about Unbiased Estimator

• An estimator is unbiased when the mean of its sampling distribution equals the population parameter it estimates.

• Unbiased does not mean every estimate is correct; it means the estimates do not systematically overshoot or undershoot.

• The sample mean x̄ is unbiased for μ, and the sample proportion p̂ is unbiased for p when the sample is random.

• The sample variance s² is unbiased for σ² because it divides by n − 1, while dividing by n produces a biased estimator that tends to be too small.

• Bias and variability are separate qualities, and increasing sample size shrinks variability but does nothing to fix bias.

• To judge bias from a simulation, look at where the sampling distribution is centered compared to the true parameter value.

Frequently asked questions about Unbiased Estimator

What is an unbiased estimator in AP Stats?

It's a statistic whose sampling distribution is centered at the true parameter, so its average value across all possible random samples equals that parameter. The standard examples are x̄ for μ and p̂ for p.

Does an unbiased estimator always give the right answer?

No. A single estimate from an unbiased estimator can still be far from the parameter because of sampling variability. Unbiased only means the estimates are correct on average and not off in one consistent direction.

Does a larger sample size make a biased estimator unbiased?

No. A larger sample shrinks the spread of the sampling distribution, but if the distribution is centered in the wrong place, it stays centered in the wrong place. To reduce bias, you need a better estimator or a better sampling method.

Why do we divide by n − 1 for sample variance?

Dividing the sum of squared deviations by n tends to underestimate σ², because deviations are measured from x̄ instead of the true μ. Dividing by n − 1 corrects for this, which makes s² an unbiased estimator of σ².

How is an unbiased estimator different from a biased estimator?

An unbiased estimator's sampling distribution is centered on the parameter. A biased estimator's sampling distribution is centered somewhere else, so it consistently overestimates or underestimates. On a simulated dotplot, you can spot the difference by comparing the center of the dots to the true parameter value.