Selection Bias
Occurs when the sample isn’t representative of the population due to non-random selection (e.g., only surveying tech-savvy people about internet habits).
Undercoverage Bias
A specific kind of selection bias — parts of the population are not included or underrepresented (e.g., excluding rural areas from a survey).
Survivorship Bias
Only analyzing “survivors” or successful cases and ignoring those that failed or dropped out (e.g., studying only successful startups and ignoring the failed ones).
Sampling Bias
General term for any bias introduced due to the way samples are collected — includes both selection and undercoverage bias.
Nonresponse Bias
Happens when individuals selected for a sample don’t respond, and their nonresponse is related to the variable of interest (e.g., dissatisfied customers not filling out a survey).
Measurement Bias (or Instrument Bias)
Occurs when the measurement tool skews the data (e.g., faulty sensor, poorly worded survey questions).
Recall Bias
Especially common in surveys — people may not remember past events accurately (e.g., asking patients to recall dietary habits from 5 years ago).
Observer Bias (Experimenter Bias)
When the researcher’s expectations subtly influence what they observe or record.
Confirmation Bias
Focusing on data that confirms existing beliefs, ignoring contradictory evidence.
Reporting Bias
Only certain results are shared — often positive ones — skewing the perception of effectiveness or truth (e.g., in clinical trials or marketing data).
Omitted Variable Bias
Leaving out a relevant variable from a model that affects both the predictor and the outcome can distort results (e.g., ignoring education when modeling income).
Algorithmic Bias
Bias introduced by models trained on biased data, often leading to unfair or discriminatory outcomes (e.g., facial recognition performing worse on certain skin tones)