Bias

In empirical work, any systematic difference between the empirical results of an analysis and the true facts of the case (for example, the difference between

Binomial Distribution 31

the distribution of values in a sample and the actual values of the population from which the sample is drawn). In non-statistical areas it is any distorting influence which might lead to wrong or misleading results, for example, a search of the (English language) literature on a subject might lead one to ignore all Chinese contributions (unfortunately, no reviewer knew Chinese) and to conclude something wrong about the results (apart from the apparent fact that the Chinese were not in the field). Research sponsorship (whether by commercial or non-commercial sponsors) can lead to pressure on researchers to produce particular results or suppress 'unwanted' results. Common types of bias in clinical trials and surveys include allocation bias, ascertainment bias, design bias, detection bias, exclusion bias, information bias, interviewer bias, lead-time bias, measurement bias, observer bias, performance bias, publication bias, recall bias, referral bias, sample selection bias, selection bias, surveillance bias, therapeutic personality bias, volunteer bias, withdrawal bias and work-up bias. See also End of Scale Bias, Omitted Variable Bias, Spacing out Bias, Starting Point Bias.

Was this article helpful?

0 0

Post a comment