Definition
An estimator is a rule that maps observed data to an estimate of an unknown parameter or functional. The rule is random before the data are observed, so its sampling distribution matters. Bias, variance, consistency and robustness describe different properties of that rule.
How it is calculated
An estimator theta-hat is a statistic T applied to the observed sample X_1 through X_n.
| Symbol | Meaning | Unit |
|---|---|---|
| theta | unknown target parameter | target units |
| T | estimation rule | mapping |
| X | observed data | data units |
| n | sample size | count |
Interpretation
An estimator is the measuring instrument, not the measurement itself. Before seeing data it has a distribution; after seeing data it produces one number. Two estimators can target the same quantity while trading bias against variance.
Assumptions and conventions
-
The target parameter and data-generating setup are defined.
-
The estimator's domain includes the observed sample.
The estimator has the same units as the target parameter.
Limitations and common misreadings
-
An estimate is not the same as the unknown parameter.
-
Unbiasedness alone does not make an estimator precise.
References
- Larry Wasserman. (2004). All of Statistics: A Concise Course in Statistical Inference. Springer Texts in Statistics. Chapter 6, Models, Statistical Inference and Learning. doi:10.1007/978-0-387-21736-9. Source

