Unbiasedness and consistency

Understanding Unbiasedness and Consistency in Estimators

Key Properties of Estimators

• The discussion focuses on two important properties of estimators: unbiasedness and consistency, which are often confused in statistical contexts.

• Unbiasedness means that the expected value of an estimator (e.g., β̂) equals the true population parameter. This property ensures that, on average, the estimator accurately reflects the population value.

• Consistency indicates that as sample size increases (n → ∞), the estimated value approaches the true population parameter. This is a desirable trait because it implies reliability with larger samples.

Examples of Estimators

• An example provided illustrates an estimator (β̂) that is both unbiased and consistent, highlighting its effectiveness in estimating population parameters accurately across different samples.

• A contrasting example introduces a biased but consistent estimator (β̃). Although its expected value may be higher than the actual population parameter, it can still converge to this parameter as sample size increases.

Implications of Bias and Consistency

• The concept of bias is explored further; for instance, if β̃ consistently overestimates a parameter, it is termed upwardly biased. However, increasing sample sizes can lead to more accurate estimates over time.

• The speaker emphasizes that while having an unbiased and consistent estimator is ideal, sometimes one must work with biased yet consistent estimators when necessary.

Limitations of Non-representative Estimators

• Discussing non-useful estimators, such as one that outputs a constant value regardless of input (e.g., always returning 4), highlights their lack of representativeness concerning actual population parameters.