Skip to content
AP Statistics · Inference for Categorical Data: Proportions

Chi square

Chi square appears in 43 reviewed AP Statistics questions and sits in Inference for Categorical Data: Proportions. Below are worked examples showing exactly how the exam tests it, with the answer marked and every distractor explained.

How Chi square is tested

Example 1Easier
In a goodness-of-fit test, the null hypothesis states that births are equally distributed across all 7 days of the week. A hospital records 700 births. What is the expected count per day?
  1. A700
  2. B50
  3. C100Correct
  4. D7

Why C is right

Under H₀ of equal distribution, each day should have the same expected count: E = n/k = 700/7 = 100 births per day. The proportion-based inference follows from the sample size, observed count, and standard error in the problem.
Why the other 3 choices miss
A
Incorrect because 700 is the total sample size, not the expected count for any individual category.
B
Incorrect because 50 would be the expected count per category if n=350 and k=7, or n=700 and k=14, neither of which applies here.
D
Incorrect because 7 is the number of categories (days of the week), not the expected count per category.
Example 2Medium
A researcher compares the distribution of political party preference (Democrat, Republican, Independent) between two cities using a chi-square test. Both cities are sampled independently. This test is a chi-square test for
  1. Ahomogeneity, because two independent populations are compared on one variableCorrect
  2. Bindependence, because party preference and city are both measured on the same individuals
  3. Cgoodness of fit, because party preference has three categories in each city sample
  4. Dassociation, because the test determines causation between city and party preference

Why A is right

A chi-square test for homogeneity compares the distribution of a categorical variable across two or more independently sampled populations. Here, the two cities are separate populations, and party preference is the categorical variable being compared. The question is whether the distribution of party preference is the same (homogeneous) in both cities.
Why the other 3 choices miss
B
Incorrect. A test for independence examines whether two variables measured on the same individuals are associated. Here, the two cities are separate samples, not two variables measured on the same people.
C
Incorrect. A goodness-of-fit test compares one sample's distribution to a hypothesized distribution. Having three categories does not make it goodness of fit; this test involves two samples, not one sample compared to a theoretical distribution.
D
Incorrect. Chi-square tests assess association, not causation. Furthermore, the correct classification is homogeneity (comparing distributions across populations), not simply 'association.'
Example 3Harder
A researcher finds a statistically significant chi-square test for independence (p < 0.001) between gender and career preference in a sample of 10,000 adults. However, Cramer's V (a measure of effect size for categorical variables) is only 0.04. This result illustrates that
  1. Athe sample size of 10,000 is too large for a valid chi-square test of independence
  2. Bthe chi-square test must be incorrect because the reported effect size is so small
  3. Cwith large samples, trivial associations can still be statistically significantCorrect
  4. Dgender and career preference are strongly related because the p-value is below 0.001

Why C is right

With very large samples, chi-square tests have enormous power and can detect even tiny departures from independence. A Cramer's V of 0.04 indicates an extremely weak association that, while statistically significant (unlikely due to chance), has virtually no practical importance. This demonstrates the critical distinction between statistical significance and practical significance.
Why the other 3 choices miss
A
Incorrect. There is no upper limit on sample size for chi-square tests. The test is valid; the issue is interpreting a significant but practically meaningless result.
B
Incorrect. The test is not incorrect. The tiny effect truly exists in the sample; it is just too small to be meaningful. Statistical significance correctly identifies a non-zero association.
D
Incorrect. A very small p-value indicates that the association is unlikely to be due to chance, but it says nothing about the strength of the association. Cramer's V = 0.04 shows the association is very weak.

Studied alongside

Other concepts from Inference for Categorical Data: Proportions.