Statistical bias

From Ganfyd

Jump to: navigation, search

The following sources of bias apply to any type of study.

There are three main types: selection, information, and confounding bias.

Sampling bias 
Ascertainment bias 
A form of sampling bias
Volunteer bias 
Recall bias
Recording bias
Response bias
Beliefs of observer (assessment bias) or subject - including blinding 
Drop-out 
Publication bias
Length-time bias 
Less aggressive, slow-growing cancers are present for longer and are more likely to be discovered by screening. Cancers detected by screening may therefore appear to do better because aggressive, fast-growing tumours are less represented.
Lead-time bias 
An artefactual improvement in survival by simply diagnosing the condition earlier. For instance, an incidental CT may detect an asymptomatic, but incurable pancreatic cancer. Well defined in Clinical Epidemiology.
Instrument validity 
especially interviews and questionnaires.
Personal tools