Search by Outline Set Search Limits Advanced Search Back Home

Selection Bias in Statistical Analysis

    Background
    • This is a common type of systematic error that occurs in studies where there is:
      • Inappropriate selection of patients
      • Unrepresentative sample of the target population
      • Insufficient retention of subjects
    Common Types of Selection Bias
    • Attrition Bias:
      • Occurs when there is a high loss of study subjects within a study (eg., lost to follow up) and the remaining sample now differs from the original sample being studied that was supposed to be representative of the target population.
    • Berkson Bias:
      • The disease being studied is only being done from patients in the hospital that may not lead to results representative of the target population.
    • Neyman Bias:
      • Occurs when the exposure occurs before the assessment of the disease resulting in missed cases that die early or happen to recover.
    • Nonresponse Bias:
      • Occurs when this is a poor response rate of the desired sample population where the responders now are different from the nonresponders.
    • Sampling Bias:
      • Occurs when a study does not randomize patients at all or does not appropriately randomize subjects to where there the study population does not represent the target population.
    Considerations
    • If selection bias is present it can lead to differences in the measure of association between the sample versus target population. 
      • For example, differences can occur with the incidence rate or odds ratio and thus is not reflective of the true difference between the groups studied and the target population. 
    • This type of bias occurs often with surveys where a high nonresponse rate leading to a possible difference between those who fill out the survey from the majority who did not.