selection bias

selection bias2016-09-19T00:56:41+00:00

selection bias 


This term has been defined differently depending on the perspective of the writer, e.g., as a trialist, an epidemiologist, a public health worker, and so on. I’d rather not consolidate the below definitions, so here the reader can decide which best meets his or her needs

(1) The inadvertent selection of a non-representative sample of subjects or observations

(2) Any bias (error) in a data set introduced by the investigator, which results in a directional deviation from randomness–e.g., differences in characteristics of participants selected for a study compared to those not selected, or selection of data for inclusion in a final analysis

(3) A bias in assignment or a confounding variable that arises from study design rather than by chance

Prevention Random allocation with adequate concealment of assignment 

Selection bias can occur when the study and control groups are chosen so that they differ from each other by one or more factors that may affect the outcome of the study  

(4) A systematic error in reviews selected for inclusion in a journal article

Example Publication bias

(5) A systematic difference in characteristics between subjects selected for study and those who are not, which affects the generalisability (external validity) of a study but not its (internal) validity

A classic example of selection bias is a 1936 poll which predicted President Roosevelt not getting re-elected because telephone directories were used to select respondents, and most people didn’t have phones at the time

Synonyms Sampling bias, spectrum bias 



Leave A Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.