Group comparison

Group comparison capabilities let you check whether the average of a numeric feature differs between two or more groups, telling you how large the difference is and how confident you can be that it's real. Choose the method that matches your comparison:

• T-test: two groups.
• ANOVA: three or more groups.

T-test

The two-sample t-test determines whether the mean of a feature differs between two groups.

1. Open a table.
2. Run Top Menu > ML > Analyze > Group Comparison > T-test.... A dialog opens.
3. In the dialog, specify:
   • the column defining the two groups (in the Category field)
   • the column with feature values (in the Feature field)
   • the significance level (in the Alpha field)
   • the analysis method (in the Method field): Welch or Student
4. Click Run to execute. A box plot and an Analysis/Conclusion tab control appear.

Datagrok supports two two-sample t-test methods:

• Welch (default) - robust to unequal variances across groups. Recommended unless you have strong reason to assume equal variances.
• Student - classical t-test. More powerful when variances are equal, but unreliable otherwise. You can't run the analysis if group variances differ significantly - switch to Welch in that case.

The box plot shows the distribution of values by categories:

The Analysis tab reports the t-statistic, degrees of freedom, p-value, mean difference with its confidence interval, and effect size (Cohen's d and Hedges' g). The Conclusion tab presents the null hypothesis testing.

ANOVA

Analysis of variance (ANOVA) determines whether the examined factor has a significant impact on the studied feature.

1. Open a table.
2. Run Top Menu > ML > Analyze > Group Comparison > ANOVA.... A dialog opens.
3. In the dialog, specify:
   • the column with factor values (in the Category field)
   • the column with feature values (in the Feature field)
   • the analysis method (in the Method field): Welch or Fisher
   • the significance level (in the Alpha field)
4. Click Run to execute. The following analysis appears:

Datagrok supports two one-way ANOVA methods:

• Welch (default) - robust to unequal variances across groups. Recommended unless you have strong reason to assume equal variances.
• Fisher - classical ANOVA. More powerful when variances are equal, but unreliable otherwise. You can't run the analysis if group variances differ significantly - switch to Welch in that case.

The Analysis tab presents a table with ANOVA computations:

The Fisher and Welch methods show different columns:

• Fisher: sums of squares (SS), degrees of freedom (DF), mean squares (MS), F-statistic, critical F-value, and p-value - split into Between groups, Within groups, and Total.
• Welch: F-statistic, numerator df (k − 1), Welch–Satterthwaite denominator df (fractional), critical F-value, and p-value - Welch's test has no SS/MS decomposition by design.

Click the Conclusion tab to explore the null hypothesis testing.

See also:

• Box plot