GLRC 2026
Participate Home Rounds Awards Training Blog FAQ
Resources Resources Support

Base rate fallacy simple example with 100 people

Blog › Logic And Reasoning Explained

· by Julia Berens · 4 min read

A base rate fallacy simple example starts with 100 people and a fictional screening test that catches every case of a condition. Only four people in the group have the condition, but 12 people test positive. If you pick one person at random from those 12, the chance that they have the condition is one in three. The test did not miss anyone with the condition. It just flagged some people without it, too.

Count the 100 people before reading the results

Before anyone takes the test, split the group into two rows: four people with the condition and 96 without it. The base rate is the starting frequency of the condition in the group: 4 out of 100, or 4%. It tells you how common the condition is before you know anybody's test result.

For this exercise, assume the test detects everyone who has the condition. It also gives a positive result to exactly 1 in 12 people who do not have the condition. The numbers are made up to keep the arithmetic simple. Since there are 96 people without the condition, 96 divided by 12 gives eight positive results in that row.

GroupPositiveNegativeTotal
Has the condition404
Does not have the condition88896
Total1288100

The four positive results in the first row are true positives. The eight in the second row are false positives. A false positive is still a positive test result. That is why all eight belong in the positive column, even though none of those eight people has the condition.

Working through the base rate fallacy simple example

Compare these two questions. They both use the same four people in the top left cell of the table, but they divide by different totals.

  • What is the chance of testing positive if you have the condition? Look only at the four people who have it. All four test positive, so the answer is 4 out of 4, or 100%. This is the test's sensitivity in the example.
  • What is the chance of having the condition if you test positive? Look only at the 12 people who test positive. Four have the condition, so the answer is 4 out of 12, or one third. This is the positive predictive value in the example.

The words after if tell you which group to start with. In the first question, that group is a row. In the second, it is a column. Treating 100% sensitivity as a 100% chance of having the condition after a positive result reverses the two probabilities. It also ignores the base rate: the group without the condition is so much larger that even a modest share of false positives outnumbers the true positives.

That does not make the result meaningless. Before seeing a test result, a person picked at random from the whole group has a 4% chance of having the condition. Among those with a positive result, the share is about 33%. The positive result changes what you know, but it does not settle the question.

The same test in a group with a higher base rate

Keep the group size at 100 and the test performance unchanged. This time, suppose 28 people have the condition and 72 do not. The test still catches every case, so there are 28 true positives. It still flags 1 in 12 people without the condition, so there are six false positives.

Now there are 34 positive results. Of those, 28 come from people with the condition. That is 28 out of 34, or about 82%, compared with about 33% in the first group. The test did not get better at detecting the condition. It also did not become less likely to flag a person without it. What changed was the number of people in each starting group.

This is why even a correctly reported positive predictive value needs context. A value worked out for one group does not automatically apply to another group. In these two fictional groups, the test behaves the same way within each row, yet the mix of people in the positive column changes.

Check the denominator before trusting a percentage

If a headline says a test is “92% accurate,” do not plug 92% into a calculation until you know what was counted. For instance, 92 of the 100 results in the first group are correct: four true positives plus 88 true negatives. But if your question is about positive results, those 88 true negatives are outside the group you need. Overall accuracy uses all 100 people as its denominator. Positive predictive value uses only the 12 who tested positive.

On paper, write the group you are starting with before you write the fraction. For the second question, you could write, “Out of everyone who tested positive, count those who have the condition.” Then point to the cells that supply each number. If the information you have does not tell you how many positives came from people without the condition, sensitivity alone cannot fill that gap.

For more practice with reasoning problems, see the GLRC training and practice problems and learning resources.

Try one change to the original table: keep the four true positives but cut the false positives from eight to four. Before calculating, predict which fraction will stay the same. Sensitivity is still 4 out of 4, but now 4 out of 8 positive results are true positives. Only one of the two denominators changed.

Frequently asked questions

What is a base rate?
It is how common something is in the group before you use new evidence, such as a test result. In the first example, 4 of the 100 people have the condition, so the base rate is 4%.

How can a test catch every case but still give false positives?
Catching every case describes what happens to people who have the condition. False positives come from a different group: people who do not have the condition.

What goes in the denominator after a positive result?
Count everyone who tested positive, including both true positives and false positives. In the first group, that is 4 plus 8, or 12 people.

About the Global Logic and Reasoning Competition: an online logic and reasoning competition for high school students from around the world. How the rounds work

Written by Julia Berens, Social Media Assistant of the GLRC team.

More articles

All articles