Chi-Square Test
AddMaple runs chi-square tests for you automatically to determine whether there is a significant relationship between two categorical columns and the strength of that relationship.
To get started simply pivot two columns together:
If the columns are categorical, then AddMaple will run the chi-square test and give you an overview of the result in the legend
Click "see more" to get more details behind the calculation.
This will take you to the stats tab which explains why we've run the test and the result.
If you scroll down you will see the numeric results:
P-Value - the measure of probability as to whether the relationship between two columns is due to chance or not
Cramér's V - the strength of the relationship between two categorical columns, giving a value from 0 (no relationship) to 1 (perfect relationship)
Chi Square Statistic - the difference between the actual counts and the counts you would expect if there were no relationship (null hypothesis) in the categorical data.
Degrees of Freedom - the number of values in a calculation that are free to vary. In the Chi-Square test, it is calculated based on the number of categories in each variable
Expected values under 5 - the percentage of expected values that are less than 5 (this should be less than 20% for the test to be accurate)
Expected values under 1 - the percentage of expected values that are less than 1 - this should be 0 for the test to be accurate (you may need to filter out categories with small numbers of results via "Options" -> "Add Filter")
Further Insights Between Column Categories
When you have more than two categories in your pivot, AddMaple provides additional analysis to identify which specific categories are driving the significant relationship. Click the toggle "Further Insights Between Column Categories" to see a detailed breakdown.
This analysis compares each category against all the other categories combined to determine which ones are significantly different from the rest. The results are ordered by the level of difference, showing you:
- Category: The specific category being analyzed
- Level of difference: Strong, Moderate, or None (based on statistical significance and effect size)
- P-value: The statistical significance of that category's difference from the rest
- V-value: The effect size (Cramér's V) showing the strength of the relationship for that specific category
For example, if you're analyzing "Job Category" vs "Satisfaction Level", this feature will show you that "Manager" has a strong relationship with satisfaction while "Software Engineer" shows no significant difference from the overall pattern.
This helps you understand not just that there's a relationship between your columns, but which specific categories are most responsible for driving that relationship.