Back to Audiences

Statistical Analysis

Compare metrics across customer groups. Levene's test runs automatically to pick the right t-test.

No numeric data available on your customers (0 loaded, all with empty metadata).

To run t-tests you need numeric fields (income, spend, age, score…). Re-import your CSV and map those columns — unknown columns are now automatically captured into metadata.

Go to Import →

Grouping variable

Group A

Group B

Numeric metric

No numeric metadata fields found on customers — add numeric values to enable t-tests.

function levenesTest(groupA, groupB) { const meanA = groupA.reduce((s, x) => s + x, 0) / groupA.length; const meanB = groupB.reduce((s, x) => s + x, 0) / groupB.length; const devA = groupA.map((x) => Math.abs(x - meanA)); const devB = groupB.map((x) => Math.abs(x - meanB)); const meanDevA = devA.reduce((s, x) => s + x, 0) / devA.length; const meanDevB = devB.reduce((s, x) => s + x, 0) / devB.length; const allDevs = [...devA, ...devB]; const grandMean = allDevs.reduce((s, x) => s + x, 0) / allDevs.length; const nA = groupA.length; const nB = groupB.length; const N = nA + nB; const numerator = (N - 2) * (nA * Math.pow(meanDevA - grandMean, 2) + nB * Math.pow(meanDevB - grandMean, 2)); const denominator = devA.reduce((s, x) => s + Math.pow(x - meanDevA, 2), 0) + devB.reduce((s, x) => s + Math.pow(x - meanDevB, 2), 0); const F = denominator === 0 ? 0 : numerator / denominator; const p = 1 - jStat.centralF.cdf(F, 1, N - 2); return { F: round(F), df1: 1, df2: N - 2, p: round(p), equalVariance: p > 0.05 }; }function studentsTTest(groupA, groupB) { const nA = groupA.length; const nB = groupB.length; const meanA = groupA.reduce((s, x) => s + x, 0) / nA; const meanB = groupB.reduce((s, x) => s + x, 0) / nB; const varA = groupA.reduce((s, x) => s + Math.pow(x - meanA, 2), 0) / (nA - 1); const varB = groupB.reduce((s, x) => s + Math.pow(x - meanB, 2), 0) / (nB - 1); const pooledVar = ((nA - 1) * varA + (nB - 1) * varB) / (nA + nB - 2); const pooledSE = Math.sqrt(pooledVar * (1 / nA + 1 / nB)); const t = (meanA - meanB) / pooledSE; const df = nA + nB - 2; const p = 2 * (1 - jStat.studentt.cdf(Math.abs(t), df)); const tCritical = jStat.studentt.inv(0.975, df); const marginOfError = tCritical * pooledSE; const cohensD = Math.abs(meanA - meanB) / Math.sqrt(pooledVar); return { t: round(t), df: round(df, 2), p: round(p), meanA: round(meanA, 2), meanB: round(meanB, 2), varA: round(varA, 2), varB: round(varB, 2), pooledVar: round(pooledVar, 2), ciLower: round(meanA - meanB - marginOfError, 2), ciUpper: round(meanA - meanB + marginOfError, 2), cohensD: round(cohensD), testType: "Student's t-test", nA, nB }; }function welchsTTest(groupA, groupB) { const nA = groupA.length; const nB = groupB.length; const meanA = groupA.reduce((s, x) => s + x, 0) / nA; const meanB = groupB.reduce((s, x) => s + x, 0) / nB; const varA = groupA.reduce((s, x) => s + Math.pow(x - meanA, 2), 0) / (nA - 1); const varB = groupB.reduce((s, x) => s + Math.pow(x - meanB, 2), 0) / (nB - 1); const se = Math.sqrt(varA / nA + varB / nB); const t = (meanA - meanB) / se; const df = Math.pow(varA / nA + varB / nB, 2) / (Math.pow(varA / nA, 2) / (nA - 1) + Math.pow(varB / nB, 2) / (nB - 1)); const p = 2 * (1 - jStat.studentt.cdf(Math.abs(t), df)); const tCritical = jStat.studentt.inv(0.975, df); const marginOfError = tCritical * se; const pooledSD = Math.sqrt((varA + varB) / 2); const cohensD = Math.abs(meanA - meanB) / pooledSD; return { t: round(t), df: round(df, 2), p: round(p), meanA: round(meanA, 2), meanB: round(meanB, 2), varA: round(varA, 2), varB: round(varB, 2), pooledVar: null, ciLower: round(meanA - meanB - marginOfError, 2), ciUpper: round(meanA - meanB + marginOfError, 2), cohensD: round(cohensD), testType: "Welch's t-test", nA, nB }; }