MENTAL MODEL · M0291

Berkson's Paradox

TechnicalHigh supportStatistics

Included

Version 1.0.0 · Updated 2026-07-28

CORE DEFINITION

Berkson's paradox, also known as Berkson's bias, collider bias, or Berkson's fallacy, refers to the situation where people's intuitive observations do not match the actual conditional probabilities and rigorous statistical results; that is, the seemingly correlated factors that people discover are actually unrelated. This occurs when biased samples are included in the study design and bias exists. The paradox was proposed by American physician and statistician Joseph Berkson in 1946.

SCAFFOLDING EFFECT

psychology

Reduce cognitive load

anchor

Anchor fast decisions

For two independent events A and B with probabilities p and q respectively, in the population P(A|B)=P(A), so they are unrelated. However, if we only observe the subset where at least one of A or B occurs (e.g., patients admitted due to A or B), then in that subset the proportion of 'only A' is inflated and the proportion of 'both A and B' is deflated, causing A and B to appear negatively correlated in the subset. Essentially, this is bias due to conditioning on a collider.

MINIMUM ACTION

In progress 0/4

Practice this model in one real situation:

Confirm whether your data comes from a 'filtered' population (hospitals, award winners, paying users, interview passers)
Identify whether the filtering condition is correlated with multiple variables (i.e., it is a collider)
If you observe a negative correlation, first ask 'Is this created by the filtering?'
Re-examine with population/random sampling, or build a causal diagram to distinguish correlation from causation.

Check to track your progress (stored locally)

Learning progress0%

account_treeGenealogyexpand_more

menu_bookReferencesexpand_more

Source support: Explicit

link
zh.wikipedia.orghttps://zh.wikipedia.org/wiki/%E4%BC%AF%E5%85%8B%E6%A3%AE%E6%82%96%E8%AE%BAZH · Explicit
verified

RELATED MODELS