01 / 21

Biased & Unbiased
Point Estimates

Learn how sample statistics estimate population parameters, and discover when an estimator hits the target on average versus when it systematically misses.

AP Statistics

What is a Point Estimate?

A point estimate is a single value computed from sample data that serves as our best guess for an unknown population parameter. We use a statistic (from the sample) to estimate a parameter (of the population).

Parameter
A fixed, often unknown value describing the entire population: \(\mu\), \(\sigma\), \(p\)
Statistic
A value calculated from sample data that estimates the parameter: \(\bar{x}\), \(s\), \(\hat{p}\)

Parameters vs Statistics

Parameter Statistic
Source Population Sample
Nature Fixed (unknown) Varies by sample
Mean \(\mu\) \(\bar{x}\)
Proportion \(p\) \(\hat{p}\)
Std Dev \(\sigma\) \(s\)
Notation Greek letters Latin letters / hats

Why Do Point Estimates Matter?

In the real world, we almost never have access to the entire population. Instead, we draw a sample and use it to make inferences. The quality of our inference depends on whether our estimator is reliable and accurate.

Polling
Use \(\hat{p}\) from 1,000 voters to estimate the true proportion \(p\) who favor a candidate
Medicine
Use \(\bar{x}\) from a clinical trial to estimate the true mean treatment effect \(\mu\)
Quality Control
Use sample defect rate \(\hat{p}\) to estimate the true defect rate \(p\) in a factory

The Unbiased Estimator

An estimator \(\hat{\theta}\) is unbiased for a parameter \(\theta\) if the mean of its sampling distribution equals the parameter:

This means that if we repeated our sampling process infinitely many times, the average of all our estimates would equal the true parameter. Any single estimate might be too high or too low, but on average we hit the target.

The Biased Estimator

An estimator is biased if its expected value does not equal the parameter it estimates:

A biased estimator systematically overestimates or underestimates the true parameter, no matter how many samples you take.

Visualizing Bias vs Unbiased

True parameter \(\theta\) Unbiased: \(E(\hat{\theta}) = \theta\) Biased: \(E(\hat{\theta}) \neq \theta\) Bias

Bessel's Correction: Why \(n-1\)?

When computing sample variance, dividing by \(n\) produces a biased estimator that underestimates \(\sigma^2\). Dividing by \(n-1\) corrects this:

Biased (divide by \(n\))
$$ \tilde{s}^2 = \frac{\sum(x_i - \bar{x})^2}{n} $$ \( E(\tilde{s}^2) = \frac{n-1}{n}\sigma^2 < \sigma^2 \)
Unbiased (divide by \(n-1\))
$$ s^2 = \frac{\sum(x_i - \bar{x})^2}{n-1} $$ \( E(s^2) = \sigma^2 \) ✓

Bias vs Variability

Bias and variability are two different problems with estimators. A good estimator should have low bias and low variability.

Low Bias Low Variability Low Bias High Variability High Bias Low Variability High Bias High Variability Low Variability High Variability

Common Unbiased Estimators

1
Sample mean \(\bar{x}\) is unbiased for the population mean \(\mu\): \(E(\bar{x}) = \mu\)
2
Sample proportion \(\hat{p}\) is unbiased for the population proportion \(p\): \(E(\hat{p}) = p\)
3
Sample variance \(s^2\) (with \(n-1\)) is unbiased for population variance \(\sigma^2\): \(E(s^2) = \sigma^2\)
4
Sample slope \(b_1\) in regression is unbiased for the true slope \(\beta_1\): \(E(b_1) = \beta_1\)

How to Check for Bias

1
Identify the parameter you are trying to estimate (e.g., \(\mu\), \(p\), \(\sigma^2\))
2
Identify the statistic being used as the estimator (e.g., \(\bar{x}\), \(\hat{p}\), \(s^2\))
3
Find the expected value of the statistic: compute \(E(\hat{\theta})\)
4
Compare: If \(E(\hat{\theta}) = \theta\), the estimator is unbiased. If \(E(\hat{\theta}) \neq \theta\), it is biased.
5
Calculate the bias if needed: \(\text{Bias} = E(\hat{\theta}) - \theta\). Positive bias means overestimation; negative means underestimation.

Example 1: Sample Mean

A researcher measures the heights of 25 randomly selected students and gets \(\bar{x} = 67.3\) inches. The true mean height of all students is \(\mu\).

Example 2: Sample Proportion

In a poll of 500 voters, 280 say they support Candidate A. What is the point estimate of the true proportion \(p\) who support Candidate A?

Example 3: Biased Variance Estimator

Suppose we have \(n = 5\) data points and we compute variance using \(\frac{1}{n}\) instead of \(\frac{1}{n-1}\). How large is the bias?

Common Mistakes

✕ Confusing Bias with Variability
Bias means the estimator is systematically off-center. High variability means estimates are spread out. These are separate concepts.
✕ Using \(n\) Instead of \(n-1\)
Always use \(n-1\) in the denominator for sample variance \(s^2\) to get an unbiased estimate of \(\sigma^2\).
✕ Thinking "Unbiased" Means Correct
An unbiased estimator is correct on average, not for any single sample. Each individual estimate may still miss the parameter.

Summary of Key Estimators

Estimator Target Parameter Unbiased?
\(\bar{x}\) \(\mu\) ✓ Yes
\(\hat{p}\) \(p\) ✓ Yes
\(s^2\) with \(n-1\) \(\sigma^2\) ✓ Yes
\(\tilde{s}^2\) with \(n\) \(\sigma^2\) ✕ No (underestimates)
\(s\) \(\sigma\) ✕ No (slightly biased)
\(b_1\) (regression slope) \(\beta_1\) ✓ Yes

Question 1 of 5

An estimator \(\hat{\theta}\) is unbiased if:

Question 2 of 5

Which of the following is a biased estimator of its corresponding parameter?

Question 3 of 5

A sample of \(n = 10\) observations is used to compute variance with \(\frac{1}{n}\). The expected value of this estimator equals:

Question 4 of 5

In a survey, 120 out of 400 respondents prefer Brand X. The point estimate of the true proportion \(p\) who prefer Brand X is:

Question 5 of 5

If an estimator has low bias but high variability, what does this mean?