An unbiased estimator is a sample statistic whose sampling distribution is centered at the population parameter it estimates, meaning that across all possible samples, the average value of the statistic equals the true parameter (AP Stats Topic 5.4).
An unbiased estimator is a statistic that doesn't systematically miss. If you took every possible random sample, calculated the statistic for each one, and averaged all those values, you'd land exactly on the true population parameter. Individual estimates still bounce around (that's sampling variability), but they bounce around the right target. Picture a dartboard. An unbiased estimator's darts scatter, but they scatter evenly around the bullseye instead of clustering off to one side.
In AP Stats language, an estimator is unbiased when the mean of its sampling distribution equals the parameter. The big three you'll use all year are unbiased under random sampling. The sample mean x̄ is unbiased for μ, the sample proportion p̂ is unbiased for p, and the sample variance s² (the one with n - 1 in the denominator) is unbiased for σ². That n - 1 isn't arbitrary. Dividing by n would systematically underestimate σ², so the formula corrects for it. Unbiased does not mean accurate on any single sample. It means no systematic lean in either direction.
This term lives in Topic 5.4 (Biased and Unbiased Point Estimates) in Unit 5: Sampling Distributions. Learning objective 5.4.A asks you to explain why an estimator is or is not unbiased, and 5.4.B has you calculate point estimates while recognizing that estimators vary from sample to sample. Unit 5 is the bridge of the whole course. Everything in Units 6-9 (confidence intervals and significance tests) only works because x̄ and p̂ are unbiased estimators with predictable sampling distributions. When you write a confidence interval centered at x̄, you're trusting that x̄ isn't systematically high or low. Topic 5.4 is where you earn that trust.
Keep studying AP Statistics Unit kieImJKumIjyX99J
Sampling Distribution (Unit 5)
Unbiasedness is a property of the sampling distribution, not of one sample. An estimator is unbiased exactly when its sampling distribution is centered at the parameter. You can't even define bias without imagining the statistic computed over many, many samples.
Point Estimate (Unit 5)
A point estimate is the single number you get from your one sample, like x̄ = 14.2. Unbiasedness is the quality-control label on the recipe that produced that number. The estimate is the output; the estimator is the method, and only the method can be biased or unbiased.
Confidence Interval (Units 6-8)
Every confidence interval has the form (unbiased) point estimate ± margin of error. If the estimator were biased, the interval would be built around the wrong center and the stated confidence level would be a lie. Unbiasedness is the quiet assumption holding all of inference together.
Bias in Sampling Methods (Unit 3)
Unit 3 bias (nonresponse, undercoverage, voluntary response) comes from how data is collected. Unit 5 bias comes from the math of the estimator itself. Even a perfectly unbiased estimator like x̄ gives systematically wrong answers if the sample wasn't random in the first place.
This shows up as multiple choice that tests whether you can explain unbiasedness, not just recite it. Classic stems ask why x̄ is an unbiased estimator of μ (correct answer: the mean of the sampling distribution of x̄ equals μ), or why s² with the n - 1 divisor is unbiased while dividing by n is not. Another common angle shows you a dotplot of a simulated sampling distribution and asks whether the estimator appears biased, which you answer by comparing the center of the distribution to the true parameter. No released FRQ has used the phrase verbatim, but investigative-task FRQs love handing you an unfamiliar estimator and asking whether it's biased based on simulation results. The winning move is always the same sentence: compare the average value of the estimator across samples to the parameter. Saying "the estimate equals the parameter" without the "on average, across all samples" framing loses the point.
A biased estimator's sampling distribution is centered somewhere other than the true parameter, so it systematically over- or underestimates. The classic example is dividing by n instead of n - 1 when computing sample variance, which consistently underestimates σ². The trap is thinking unbiased means low variability. An estimator can be unbiased but wildly variable, or biased but very consistent. Bias is about where the distribution is centered; variability is about how spread out it is. They're separate dials.
An estimator is unbiased if the mean of its sampling distribution equals the population parameter it's estimating.
Unbiased describes the long-run average behavior of the method across all samples, not the accuracy of any single estimate.
Under random sampling, x̄ is unbiased for μ, p̂ is unbiased for p, and s² is unbiased for σ².
The sample variance formula divides by n - 1 instead of n precisely because dividing by n would systematically underestimate σ².
Bias and variability are different things; an estimator can be unbiased with high variability or biased with low variability.
To judge bias from a simulated sampling distribution, compare the center of the simulated values to the true parameter value.
An unbiased estimator is a statistic whose sampling distribution is centered at the population parameter, so on average across all possible samples it equals the true value. The sample mean x̄ and sample proportion p̂ are the standard examples (Topic 5.4).
No. Any single sample will almost certainly miss the parameter because of sampling variability. Unbiased only means the misses don't lean one direction; over many samples, the high and low estimates average out to the true value.
Dividing by n produces a biased estimator that systematically underestimates the population variance σ², because deviations are measured from x̄ rather than the true mean μ. Using n - 1 corrects that, which is exactly why s² is the unbiased estimator AP Stats uses.
A biased estimator is a math problem; the formula itself is centered at the wrong value, like dividing by n for variance. Sampling bias (Unit 3) is a data-collection problem, like nonresponse or undercoverage. An unbiased formula applied to a biased sample still gives systematically wrong answers.
Find the mean of the simulated sampling distribution and compare it to the known parameter value. If the average of the simulated statistics is approximately equal to the parameter, the estimator appears unbiased; if it's noticeably off-center, it appears biased.