STATISTICAL BIAS
You want to know what voters think. You only poll people who answer their phones. The results are wrong. That is selection bias.
QUICK ANSWER
Selection bias is a bias introduced when the sample you study is not representative of the population you want to understand. It occurs when the selection process systematically excludes or includes certain groups. The bias is common in research, polling, and data analysis. It leads to misleading conclusions.
Selection bias is simple: your sample is not representative. You study one group. You think you are studying everyone. You are wrong.
If you poll only people who answer their phones, you miss people who do not answer. The results are biased.
When reading a study, ask: how was the sample selected? Is it representative?
CORE IDEA
Selection bias is simple: your sample is not representative. You study one group. You think you are studying everyone. You are wrong.
If you poll only people who answer their phones, you miss people who do not answer. The results are biased.
The solution is to ensure your sample is representative. Random sampling is the best way.
Selection bias has been recognized for centuries. In the 1930s, pollsters learned the lesson the hard way. The Literary Digest polled its readers to predict the 1936 US presidential election. The poll predicted a landslide victory for Alf Landon. Franklin Roosevelt won in a landslide. The poll was wrong because it only polled people who subscribed to the magazine. Those people were not representative.
The lesson is simple: the sample must be representative. If it is not, the results are meaningless.
Today, selection bias is a foundational concept in statistics and research. It is taught in every statistics class.
Selection bias became famous because of the Literary Digest poll. The poll was a spectacular failure. It showed the importance of representative sampling.
The concept is central to statistics and research. It explains why many studies are flawed.
Today, selection bias is a foundational concept in statistics. It is taught in every statistics class.
The Literary Digest poll of 1936 is the classic example. The poll was wrong because the sample was not representative.
A study of health only includes people who visit doctors. The results are biased. Sick people visit doctors more often.
A poll only includes people who answer their phones. The results are biased. Some people do not answer.
A company surveys its customers. It only surveys satisfied customers. The results are biased.
CONCEPT MAP
A sample misleads because the way it was selected is distorted.
Survivorship bias is a selection problem.
The statistical version of the same issue.
Bad samples hide true rates.
Visible cases are not always representative.
Regression to the Mean points back to this idea in the concept network.
No. Survivorship bias is a specific type of selection bias. Selection bias is the broader category.
No. It applies to polling, business, and everyday life. Any time you make a generalization, selection bias can occur.
You cannot eliminate it. You can only reduce it. Random sampling is the best defense.
Selection bias is a powerful concept, but it has limitations. Not every sample is biased. Some samples are representative.
The concept can be overused. Not every study is flawed. Sometimes the sample is good enough.
The concept is a guide, not a rule. Use it to think about samples, not to dismiss every study.
When reading a study, ask: how was the sample selected? Is it representative?
When conducting research, use random sampling. Random sampling reduces selection bias.
When making generalizations, ask: who am I missing? Who is not represented?
EXPLORE NEXT
Survivorship bias is a selection problem.
Survivorship bias is a selection problem.
Bad samples hide true rates.
Bad samples hide true rates.
Visible cases are not always representative.
Visible cases are not always representative.
Regression to the Mean points back to this idea in the concept network.
Regression to the Mean points back to this idea in the concept network.
Law of Small Numbers points back to this idea in the concept network.
Law of Small Numbers points back to this idea in the concept network.Your sample is not representative. You study one group and think you are studying everyone. You are wrong.
The Literary Digest poll of 1936. The sample was not representative. The poll was wrong.
Use random sampling. Ensure your sample is representative. Ask: who am I missing?
It leads to misleading conclusions. The results are wrong because the sample is wrong.