Chi Square Test CalculatorTest of Independence and Goodness of Fit, with Steps

How to Use This Chi Square Test Calculator

The tool runs three related procedures. Pick the tab that matches your question, then follow these steps.

1. Choose the test. Use Test of Independence when you have two categorical variables in a contingency table, such as treatment by outcome. Use Goodness of Fit when you have one variable to compare against a claimed distribution. Use P-value from χ² when you already know the statistic.
2. Set α. The default of 0.05 suits most coursework and research. Pick 0.01 or 0.001 for stricter claims.
3. Enter observed counts. Size the grid with the row and column steppers, or paste a table straight from Excel or Google Sheets. Rename the headers so your residuals and APA sentence read naturally.
4. Read the verdict and the p-value meter. If the white dot sits right of the teal α line, the result is significant.
5. Check effect size and residuals. Cramér’s V or Cohen’s w tells you how strong the pattern is. The heatmap shows which cells create it.
6. Review the assumption check. If too many expected counts fall below 5, follow the suggested alternative before you report.

Chi-Square Formula and Degrees of Freedom

Every chi-square test compares what you observed (O) with what the null hypothesis predicts (E). Squaring each gap removes the sign. Dividing by E scales each gap to the size of that cell. Adding the pieces gives one number that grows as the data drift from the null.

Yates’ continuity correction

The χ² distribution is continuous, but counts are not. In a 2 × 2 table, this mismatch makes the uncorrected test slightly too eager to reject. Frank Yates proposed subtracting 0.5 from each absolute gap: χ²_Y = Σ (|O − E| − 0.5)² / E. The correction is conservative, so many statisticians now prefer the uncorrected test for adequate samples and Fisher’s exact test for small ones. This calculator lets you toggle it and always shows both values in the steps.

Effect size: Cramér’s V and Cohen’s w

A p-value tells you whether an association exists, not how strong it is. With large samples, even trivial differences become significant. Report an effect size alongside every χ². (For numeric outcomes compared between two groups, the equivalent measure is Cohen’s d.)

Test of Independence vs Homogeneity vs Goodness of Fit

All three use the same formula. They differ in how the data were collected and in what the null hypothesis claims.

The arithmetic for homogeneity and independence is identical, so use the Test of Independence tab for both. Only the wording of your conclusion changes. If both of your variables are numeric rather than categorical, you need a correlation coefficient instead of a chi-square test.

Worked Example: Chi-Square Test of Independence

This illustrative dataset is the calculator’s default. A community pharmacy wants to know whether a reminder text message changes how patients collect refills. It follows 320 refill orders, half with an SMS reminder and half without, and records whether each refill was picked up on time, picked up late or not picked up.

1. Expected counts. For the top-left cell, E = 160 × 196 / 320 = 98. Both rows have 160 orders, so each row gets the same expected values: 98, 32 and 30.
2. Cell contributions. (112 − 98)² / 98 = 2.000; (28 − 32)² / 32 = 0.500; (20 − 30)² / 30 = 3.333. The second row mirrors the first, giving the same three values.
3. Sum. χ² = 2(2.000 + 0.500 + 3.333) = 11.667.
4. Degrees of freedom. df = (2 − 1)(3 − 1) = 2. The critical value at α = 0.05 is 5.991.
5. p-value. The right-tail area beyond 11.667 with 2 df is p ≈ 0.0029, well below 0.05, so you reject H₀.
6. Effect size. Cramér’s V = √(11.667 / (320 × 1)) = 0.19, a small association by Cohen’s benchmarks for a table with min(r, c) − 1 = 1.
7. Where the effect lives. The adjusted residuals for “Reminder × On time” (+3.21) and “Reminder × Not picked up” (−2.86) both exceed the Bonferroni cut-off of ±2.64 for six cells. The “Late” column (±1.12) does not.

Worked example: goodness of fit

A shop logs 180 customer visits: Monday 42, Tuesday 30, Wednesday 28, Thursday 35 and Friday 45. If visits were spread evenly, each day would expect 36. That gives χ² = 1.00 + 1.00 + 1.78 + 0.03 + 2.25 = 6.06 with df = 4, p ≈ 0.195 and Cohen’s w = 0.18. The difference is not significant, so the data are consistent with an even weekday spread. Load it with Try an example on the Goodness of Fit tab.

How to Interpret Chi-Square Results

p-value and critical value give the same decision

The p-value is the probability of a χ² at least this large if H₀ were true. If p ≤ α, reject H₀. Equivalently, if χ² ≥ the critical value for your df and α, reject H₀. The calculator shows both so you can match whichever method your course uses. Chi-square tests are right-tailed by design, because only large discrepancies count as evidence against H₀.

Effect size benchmarks

Cohen’s benchmarks for w are 0.10 (small), 0.30 (medium) and 0.50 (large). For Cramér’s V, divide them by √(min(r, c) − 1), so the bar drops as the table grows:

Post-hoc analysis with adjusted residuals

A significant χ² on a table larger than 2 × 2 says only that some association exists somewhere. Adjusted standardized residuals, (O − E) / √(E(1 − rowᵢ/N)(1 − colⱼ/N)), follow an approximately standard normal distribution under H₀. A value beyond ±1.96 flags a cell at α = 0.05. Testing every cell at once inflates false positives, so the calculator also applies a Bonferroni-adjusted cut-off based on the number of cells, as Sharpe (2015) recommends.

Chi-Square Test Assumptions (and What to Do When They Fail)

• Counts, not percentages or means. The test is built on frequencies. Converting to percentages hides the sample size, and the result becomes meaningless.
• Independent observations. Each subject appears in exactly one cell. For before-and-after data on the same people, use McNemar’s test instead.
• Mutually exclusive categories. No subject can fall into two columns.
• Adequate expected counts. Cochran’s widely used rule asks that no expected count fall below 1 and that no more than 20% fall below 5. For a 2 × 2 table, every expected count should ideally be 5 or more.

Chi-Square Critical Value Table

These right-tail critical values come from the same engine that powers the calculator. Reject H₀ when your χ² is greater than or equal to the value for your df and α. For any df not listed, use the P-value from χ² tab.

How to Report a Chi-Square Test in APA Format

APA style reports the statistic with its degrees of freedom and sample size, the exact p-value to three decimals (or p < .001), and an effect size. Italicize p and N, and drop the leading zero for values that cannot exceed 1:

The calculator writes this sentence for your own data, including the non-significant wording and the correct test name for goodness of fit. Add the residual findings in a second sentence when they matter to your question.

Common Chi-Square Mistakes to Avoid

• Entering percentages or proportions as counts. A table of percentages inflates or deflates χ² depending on the real N.
• Calling a significant result a strong one. With 10,000 respondents, V = 0.03 can reach p < .001. Always pair p with V or w.
• Using the wrong df for goodness of fit. If you estimated a parameter from the same data, such as a Poisson mean, subtract it as well: df = k − 1 − m.
• Ignoring sparse cells. Expected counts below 5 make the χ² approximation unreliable, even when the total N looks large.
• Treating repeated measures as independent. Paired yes/no data need McNemar’s test, not Pearson’s χ².
• Stopping at “significant.” On larger tables, report which cells drive the association using residuals.

Related Statistics Calculators

A chi-square test is usually one step in a larger analysis. These tools cover the steps before and after it.

Frequently Asked Questions

Methodology and Sources

How the numbers are computed. Expected counts use the marginal-product formula. The p-value is the regularized upper incomplete gamma function Q(df/2, χ²/2), evaluated with a series expansion or a Lentz continued fraction depending on the argument. Critical values come from numerically inverting that function. The likelihood-ratio statistic G² = 2 Σ O ln(O/E) is shown as a cross-check, because it reaches the same conclusion as Pearson’s χ² whenever expected counts are adequate. We verified the engine against published critical values, including 3.841 (df 1), 5.991 (df 2), 18.307 (df 10) and 10.828 (df 1, α = 0.001), all of which match to three decimals.

References

1. Pearson, K. (1900). On the criterion that a given system of deviations from the probable… Philosophical Magazine, 50(302), 157–175. doi:10.1080/14786440009463897
2. Yates, F. (1934). Contingency tables involving small numbers and the χ² test. Supplement to the Journal of the Royal Statistical Society, 1(2), 217–235. doi:10.2307/2983604
3. Cochran, W. G. (1954). Some methods for strengthening the common χ² tests. Biometrics, 10(4), 417–451. doi:10.2307/3001616
4. Sharpe, D. (2015). Your chi-square test is statistically significant: Now what? Practical Assessment, Research, and Evaluation, 20(8). UMass Open Publishing
5. Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences (2nd ed.). Lawrence Erlbaum Associates.
6. Agresti, A. (2018). An Introduction to Categorical Data Analysis (3rd ed.). Wiley.
7. Penn State Eberly College of Science. STAT 500, Lesson 8: Chi-Square Test for Independence. online.stat.psu.edu
8. NIST/SEMATECH. e-Handbook of Statistical Methods: Chi-Square Goodness-of-Fit Test. itl.nist.gov

Disclaimer: This calculator is an educational and analytical aid. Verify results used in published research, clinical decisions or regulatory submissions with validated statistical software, and consult a qualified statistician where needed.