Statistical Calculations
AddMaple's statistics engine automatically performs the appropriate statistical tests on your data. You don't choose tests manually—the engine examines column types, runs every relevant comparison, and surfaces the strongest relationships first.
For data experts, this accelerates your workflow so you can focus on interpreting results. For those less familiar with statistics, it removes the guesswork and presents plain-English summaries alongside the numbers.
Statistical tests at a glance
AddMaple compares your selected column against every other column in the dataset using one of these tests:
| Test | When AddMaple uses it | Effect size |
|---|---|---|
| Chi-Square | Categorical vs categorical | Cramér's V |
| T-Test | Categorical with 2 groups vs numeric | Cohen's d |
| ANOVA | Categorical with 3+ groups vs numeric | Eta squared (η²) |
| Kruskal-Wallis | Categorical vs numeric or ordinal, especially with small group sizes or non-normal data | Eta squared (η²) |
| Pearson correlation | Numeric vs numeric, normally distributed | Correlation coefficient (r) |
| Spearman correlation | Numeric vs numeric, not normally distributed | Rank correlation coefficient (ρ) |
How AddMaple chooses a test:
- Two categorical columns → Chi-Square
- One categorical, one numeric → T-Test (2 groups), ANOVA (3+ groups with sufficient data per group), or Kruskal-Wallis (very small groups or non-parametric cases)
- Two numeric columns → Pearson (normal data) or Spearman (non-normal data)
When you apply a weight column, AddMaple runs weighted versions of these tests so p-values and effect sizes reflect your weighting scheme.
How we calculate related columns
To automatically find related columns, AddMaple performs a series of statistical tests to identify significant associations between columns.
AddMaple has a powerful and fast stats engine. When you open a project, we analyze your dataset to look for relationships. For each column you select, we compare it against every other column with the appropriate test from the table above.
For each column pair, AddMaple calculates:
- Significance — The p-value shows whether the relationship is statistically significant (typically p < 0.05).
- Effect size — The practical strength of the relationship (Cramér's V, eta squared, Cohen's d, or a correlation coefficient). Effect size helps you judge importance, not just statistical significance.
Results are ranked by significance and effect size so the strongest relationships appear first. See Exploring related columns for more on viewing and interpreting them.
The six automatic tests
Chi-Square Test
Used when both columns contain categorical data. This test determines whether there is a significant association between the categories of the two variables.
Example: You survey people on preferred exercise type (running, swimming, cycling) and age group (under 30, 30–50, over 50). Chi-Square tests whether exercise preference is related to age group.
T-Test
Used when comparing one categorical column with exactly two categories against one numeric column. This test checks whether the two groups differ significantly on the numeric measure.
Example: You compare test scores for students who used traditional learning vs online learning. A T-Test determines whether the difference in scores is statistically significant.
ANOVA (Analysis of Variance)
Used when comparing one categorical column with three or more groups against one numeric column. ANOVA tests whether group means differ significantly.
Example: A study compares blood pressure changes across three diet groups (low-carb, low-fat, Mediterranean). ANOVA tests whether mean blood pressure differs across diets.
Kruskal-Wallis Test
Used when comparing one categorical column with one numeric or ordinal column, especially when some categories have small sample sizes or data does not assume a normal distribution. It tests whether group distributions differ significantly.
Example: An ecological study compares species diversity across conservation strategies in small habitat patches. With uneven group sizes, Kruskal-Wallis is appropriate instead of ANOVA.
Correlation Tests (Pearson's and Spearman's)
Used when both columns are numeric. Pearson's correlation applies when data is normally distributed; Spearman's applies when it is not. Both measure the strength and direction of the relationship.
Example: You analyze hours of exercise per week and cholesterol levels. Pearson or Spearman correlation tests whether the two numeric columns are significantly related.
How each test is calculated
Below is how AddMaple performs each test behind the scenes.
Chi-Square Test
For pairs of categorical columns:
- Calculate expected frequencies — Based on marginal totals for each category combination.
- Compute the Chi-Square statistic — Compare observed vs expected frequencies:
[ \chi^2 = \sum \frac{(O_i - E_i)^2}{E_i} ]
Where (O_i) is the observed frequency and (E_i) is the expected frequency.
- Determine p-value — Compare the statistic against the Chi-Square distribution with the appropriate degrees of freedom.
- Calculate Cramér's V — Measure association strength:
[ V = \sqrt{\frac{\chi^2}{n \times k}} ]
Where (\chi^2) is the Chi-Square statistic, (n) is total observations, and (k) is the smaller of (rows − 1) and (columns − 1).
Cramér's V tells you not just whether there is a significant association, but how strong it is.
ANOVA (Analysis of Variance)
For one categorical and one numeric column:
- Calculate group means — Mean of the numeric variable for each category.
- Compute variance — Within-group and between-group variance.
- Calculate F-statistic — Ratio of between-group to within-group variance.
- Determine p-value — Compare against the F-distribution.
- Calculate eta squared (η²) — Effect size:
[ \eta^2 = \frac{\text{SSB}}{\text{SST}} ]
Where SSB is sum of squares between groups and SST is sum of squares total. η² is the proportion of total variance explained by the categorical factor.
After ANOVA, AddMaple also runs Tukey's HSD to show which pairs of groups differ most. See How to run an ANOVA test.
Kruskal-Wallis Test
For categorical vs numeric/ordinal columns with small or uneven groups:
- Calculate group ranks — Rank all observations across groups; sum ranks per group.
- Compute H statistic — From squared ranks adjusted by group size and total observations.
- Determine p-value — Compare H against the chi-squared distribution.
- Calculate eta squared (η²) — For ordering related columns by effect size as well as p-value.
T-Test
For a categorical column with two groups vs a numeric column, AddMaple performs a two-sided T-Test:
- Calculate group means and standard deviations
- Compute T-statistic (difference in means divided by standard error)
- Determine degrees of freedom from sample sizes
- Compute p-value from the T-distribution
Correlation Tests (Pearson's and Spearman's)
For two numeric columns:
Pearson's correlation (normally distributed data):
- Check normality for both columns
- Compute Pearson's correlation coefficient (strength and direction of linear relationship)
Spearman's correlation (non-normal data):
- Rank both variables
- Compute Spearman's rank correlation coefficient
For both tests, AddMaple then calculates a t-score, degrees of freedom (n − 2), and p-value.
Step-by-step: viewing results in AddMaple
Rather than ask you to choose tests from a complex menu, AddMaple runs them automatically and presents results in an intuitive flow.
Step 1: Open related columns
When you select a column on the Chart Dashboard, AddMaple runs all calculations described above. Results appear in the Stats tab, ranked by relationship strength.
Click any related column name to pivot by that column and explore the relationship in detail.
Step 2: Read the relationship highlight
When you pivot two columns together, AddMaple shows a relationship highlight on the chart—a plain-English summary of strength (for example, moderate or strong) and the pattern in the data.
Step 3: Open the relationship overview
Click See more or open the Stats tab for dynamic paragraphs explaining which test ran, the result, and what it means for your specific columns.
Step 4: View the numbers
Turn on Statistical Test and Calculations to see p-values, test statistics, effect sizes, sample size, and degrees of freedom. Hover any metric for a description of what it means and how it was calculated.
Step 5: Explore further insights
For Chi-Square and ANOVA results, turn on Further Insights Between Column Categories to see which specific category comparisons drive the relationship.
Step 6: Explore visually on the Chart Dashboard
After pivoting by a single column, return to the Chart Dashboard to see all other columns pivoted against it, ordered by relationship strength. This lets you quickly scan which columns matter most.
Filter and recalculate
Apply filters to focus on specific categories or segments. Related columns and statistical tests are recalculated based on the filtered data.
For cell-level significance in pivot tables, see Significance testing.
Other statistical tools
These are separate from the automatic related-columns engine:
- Regression — Linear and logistic regression (More menu → Regression Analysis)
- Key Driver Analysis — Random Forest or Elastic Net (More menu → Find key drivers)
- Clustering — Group respondents by patterns across columns
- Net Promoter Score — Automatic on NPS-style recommendation columns
Guides for each test
- How to run a Chi-Square Test
- How to run a T-Test
- How to run an ANOVA test
- How to run a Kruskal-Wallis test
- Correlation Tests (Pearson's and Spearman's)
- Exploring Related Columns
- Significance Testing
- Weighting
Key points
- AddMaple automatically selects and runs the right test for each column pair
- Six core tests cover categorical, numeric, and mixed column types
- Significance (p-value) and effect size are both reported
- Related columns, pivots, filters, and the Stats tab are the main ways to explore results
- Weighted tests run when you set a project weight column
We are continually improving this feature—if there is something you'd like to see, please let us know.