Inferential Statistics
A/B & Hypothesis Testing
Compare two groups or operational variants with mathematical rigor.
Recipe Overview & Methodology
Evaluate whether observed differences in your metrics (e.g., latency, conversion, yield) are statistically significant or merely random variance. Automatically checks normality, computes Cohen's d effect sizes, and outputs p-values with 95% confidence intervals.
Statistical Validation Rules
- Levene's test for equality of variance (homoscedasticity vs. Welch's t-test).
- Student's independent two-sample t-statistic calculation.
- Two-tailed p-value with standard alpha = 0.05 threshold.
- Effect size (Cohen's d) for practical real-world significance.
Sample Output Metrics
t-Statistic
-4.82
Strong separation
p-Value
0.0018
Statistically significant
Effect Size (d)
0.86
Large practical impact
Confidence Interval
[22.0, 23.6]
95% bounds
Execution LogicClustey Engine
# Sample Clustey Python / SQL Execution
from scipy import stats
group_a = dataset[dataset['variant'] == 'control']['metric']
group_b = dataset[dataset['variant'] == 'treatment']['metric']
t_stat, p_val = stats.ttest_ind(group_a, group_b, equal_var=False)
effect_size = (group_b.mean() - group_a.mean()) / pooled_sd(group_a, group_b)Ready to apply this recipe to your own dataset?
Connect your dataset and generate this analysis in one click.