Bias

Revision as of 05:10, 26 October 2013

The term “bias” is frequently used in a colloquial sense equivalent to systematic error. In statistical sampling, it is mostly the “estimator bias” that is referred to when using this term. An estimator is biased if it does not approximate the true parametric value of a population when the sample size is more and more increased. It is eventually not the estimate that is biased, but the estimator. Any estimate deviates more or less from the true parametric value, but this may just be expression of residual variability also without bias. If a biased estimator is used, the bias is not reduced nor eliminated by increasing the sample size. “Selection bias” describes a procedure of sample selection where randomization is not fully applied but where subjective sample selection does possibly lead to the preferred selection of population elements with particular characteristics.