Biased Estimator

An biased estimator is one which delivers an estimate which is consistently different from the parameter to be estimated. In a more formal definition we can define that the expectation E of a biased estimator is not equal to the parameter of a population. In principle we can differentiate between two types of bias:
 

• location bias, where E(t) = τ + a, or a
• scale bias, where E(t) = b*τ.

Note that an unbiased statistic is not necessarily an accurate statistic. If the variance of a statistic is high it can still be unbiased.