Calculator Input
Enter observed frequency counts. Use categories in rows and outcomes in columns.
Example Data Table
This example compares training preference across three employee groups.
| Group | Online | Classroom | Hybrid |
|---|---|---|---|
| New Staff | 28 | 18 | 14 |
| Managers | 20 | 30 | 25 |
| Executives | 12 | 22 | 31 |
Formula Used
The calculator uses the chi square test of independence for contingency tables.
Pearson statistic: χ² = Σ (Oᵢⱼ - Eᵢⱼ)² / Eᵢⱼ
Degrees of freedom: df = (Rows - 1) × (Columns - 1)
Standardized residual: rᵢⱼ = (Oᵢⱼ - Eᵢⱼ) / √Eᵢⱼ
Cramer's V: V = √(χ² / (N × min(Rows - 1, Columns - 1)))
Likelihood ratio: G = 2Σ Oᵢⱼ ln(Oᵢⱼ / Eᵢⱼ)
How To Use This Calculator
- Select the number of rows and columns.
- Enter labels for each row and column.
- Add observed frequency counts in every cell.
- Choose the alpha level and correction option.
- Press the calculate button.
- Review the p value, expected counts, residuals, and effect size.
- Use CSV or PDF export for reports.
Understanding Chi Square Contingency Tables
What The Test Measures
A chi square contingency table test checks whether two categorical variables are independent. It compares the counts you observed with counts expected under independence. If the observed pattern is far from the expected pattern, the test statistic becomes larger. A larger statistic usually creates a smaller p value.
Why Expected Counts Matter
Expected counts are the backbone of this method. They show what each cell would contain if row categories and column categories had no relationship. The formula uses row totals, column totals, and the grand total. Small expected counts can weaken the usual approximation. That is why this calculator flags low expected values.
Reading The p Value
The p value answers a practical question. It estimates how unusual the observed table is when independence is true. When the p value is below your alpha level, the result is statistically significant. This does not prove causation. It only shows evidence of association between categories.
Using Residuals
The total chi square value is useful, but it does not show which cells drive the result. Residuals help solve that problem. Positive residuals show cells with more observations than expected. Negative residuals show cells with fewer observations than expected. Large absolute residuals deserve closer review.
Effect Size And Meaning
Cramer's V measures association strength. It is helpful because large samples can make small differences significant. Effect size adds context. A significant result with a tiny effect may have limited practical value. A moderate or strong effect may deserve more attention.
Good Practice
Use real frequency counts, not percentages. Keep categories clear and mutually exclusive. Avoid empty rows or columns. Combine sparse categories when it makes subject-matter sense. Review the chart, contributions, and residual tables together. This gives a fuller view than the p value alone.
FAQs
1. What is a chi square contingency table test?
It is a statistical test for two categorical variables. It checks whether the observed counts differ from counts expected under independence.
2. Can I use percentages in the table?
No. Use raw frequency counts. Percentages distort totals and can give wrong test statistics, expected counts, and p values.
3. What does a small p value mean?
A small p value means the observed table would be unusual if the variables were independent. It suggests an association exists.
4. What are expected counts?
Expected counts are estimated cell frequencies under the independence assumption. They come from row totals, column totals, and the grand total.
5. What is Cramer's V?
Cramer's V is an effect size. It shows the strength of association between categorical variables after adjusting for table size.
6. When should I use Yates correction?
Yates correction is often used for 2 by 2 tables. It makes the chi square statistic more conservative for small samples.
7. What do residuals show?
Residuals show which cells are above or below expectation. Large absolute residuals identify cells that drive the overall result.
8. Can this test prove causation?
No. The test can show association between categories. It cannot prove that one variable causes changes in another variable.