91Ó°ÊÓ

Estimators are statistical tools used to approximate unknown parameters in population data. An unbiased estimator is one that, on average, hits the true parameter it is trying to estimate. This means that the expected value of the estimator equals the parameter itself.

In mathematical terms, an estimator \(\hat{\theta}\) of a parameter \(\theta\) is unbiased if the expectation of \(\hat{\theta}\) is equal to \(\theta\). In our binomial distribution scenario, \(\hat{p}_1 = \frac{Y}{n}\) is an unbiased estimator of \(p\), meaning \(E[\hat{p}_1] = p\).

Unbiased estimators are highly desirable as they provide true reflection in estimation without systematic errors. However, while being unbiased is a good property, it does not necessarily assure the best accuracy in estimates for every sample size. Other properties, such as the variance of the estimator, also come into play when evaluating the quality of an estimator.

Mean Squared Error (MSE) helps in understanding the quality of an estimator. It is a widely used metric for assessing both the accuracy and consistency of an estimate. MSE is defined as the expectation of the square of the difference between the estimator and the true parameter value.

In formula terms, for an estimator \(\hat{\theta}\) of a population parameter \(\theta\), the MSE is calculated as: \[ \text{MSE}(\hat{\theta}) = E[(\hat{\theta} - \theta)^2] = \text{Var}(\hat{\theta}) + (\text{Bias}(\hat{\theta}))^2 \]

MSE consists of two components: variance and bias. Variance measures the estimator's variability, while bias measures how far off the average estimate is from the actual parameter value. Therefore, MSE takes into account not just how spread out the estimates are (variance), but also how close they generally are to the parameter (bias).

• A lower MSE signifies better estimator performance as it minimizes errors in both aspects.
• In the exercise, both \(\hat{p}_1\) and \(\hat{p}_2\) estimators are evaluated using their MSEs to see which provides a more reliable estimate of \(p\).

Estimator bias reflects a systematic error in estimation, where the expected value of an estimator is not equal to the true parameter it is estimating.

Mathematically, for an estimator \(\hat{\theta}\), the bias is calculated as: \[ \text{Bias}(\hat{\theta}) = E[\hat{\theta}] - \theta \]

An estimator’s bias indicates whether it consistently overestimates or underestimates the parameter. In the exercise with estimators \(\hat{p}_1\) and \(\hat{p}_2\), it is shown that while \(\hat{p}_1\) is unbiased — meaning it averages out to the true parameter \(p\) — \(\hat{p}_2\) possesses some bias as calculated by \[ \text{Bias}(\hat{p}_2) = \frac{1-p(n+2)}{n+2} \]

When evaluating different estimators, considering bias helps understand potential systematic deviations that can affect the utility of the estimator in practical applications. Often, a trade-off happens between reducing bias and controlling the variance, called the bias-variance tradeoff, which is central to the optimization of estimation methods.