Expected Frequencies and Assumptions
What You'll Learn
You'll learn how to calculate the expected frequencies for chi-squared tests and verify that your sample meets the minimum requirements for reliable results.
Why Expected Frequencies Matter
In chi-squared tests (goodness-of-fit, independence, or homogeneity), you compare observed frequencies (what you actually saw in your data) against expected frequencies (what you'd expect under the null hypothesis). The chi-squared statistic measures the discrepancy between these two. But for this test to work properly, your expected frequencies must meet certain minimum requirements—otherwise, the chi-squared distribution doesn't approximate your test statistic well.
Calculating Expected Frequencies
For a chi-squared test for independence (or homogeneity) in a two-way table:
Expected frequency for cell (i,j) = (Row i total × Column j total) / Grand total
For a goodness-of-fit test:
Expected frequency for category i = n × p₀(i)
where n is your total sample size and p₀(i) is the hypothesized probability for category i.
The Minimum Requirements
The chi-squared approximation is valid when:
- All expected frequencies ≥ 1
- At least 80% of expected frequencies ≥ 5
If these conditions fail, your p-values may be inaccurate. Think of it like needing enough "data weight" in each cell—too little, and the mathematical approximation breaks down.
Solutions when assumptions fail: combine categories (if sensible), collect more data, or use Fisher's exact test (for 2×2 tables) or other exact methods.
Key Takeaway: Always calculate expected frequencies before running a chi-squared test; verify that all are ≥1 and most (≥80%) are ≥5 to ensure the test's validity.