Cochran's C Test

Is ONE group's variance too big?

Most "equal spread" checks — Bartlett, Levene — answer a blunt question: are all the
variances equal? Cochran's C asks the sharper one that quality engineers and round-robin
labs actually care about: is a SINGLE group's variance too large for the rest? It is the
classic test for hunting down one erratic batch, machine, sensor or laboratory.

Cochran's C Test is a clean, fully OFFLINE teaching-and-calculator app. No accounts, no
data leaves your phone, and every formula is rendered right on the device.

WHAT IT DOES
• Enter 2–6 groups of numbers — thousand-separator friendly, with +/- steppers and manual
typing, add or remove a value with a tap.
• Computes each group's variance, takes the largest as a share of the total (Cochran's C),
and tells you whether that group is a genuine outlier.
• Reports the exact p-value and the 5% critical value, and warns you when the group sizes
are unequal (Cochran's C assumes a balanced design).

SEE THE RESULT, DON'T JUST READ IT
• Variance-share bar — one stacked bar split by each group's share of the total variance;
C is the width of the biggest slice.
• Variance bars — every group's variance side by side, the suspect painted in rust.
• C gauge — C on a 0..1 track with the "all equal" 1/k mark and the critical line, so you
can see at a glance how far past the threshold the suspect sits.

FIVE TABS
• Idea — what an outlying variance means and why one bad apple matters.
• Test — the calculator: verdict, charts, per-group table and the key numbers.
• Chart — the variance-share bar, the variance bars and the C gauge together.
• Sim — keep four groups with the SAME mean and inflate one group's spread; watch its slice
grow until Cochran's C flags it.
• Cases — worked examples: erratic machines, well-behaved labs, a noisy sensor.

THE MATH (rendered offline, never stacked or cramped)
• C = s²max / Σ s²i. Equal spreads give C ≈ 1/k; one group hogging the variability pushes
C toward 1.
• The critical value uses the F distribution with a Bonferroni factor k, and the engine
reproduces the published Cochran tables exactly.

WHO IT'S FOR
Quality control and Six Sigma practitioners, analytical chemists running inter-laboratory
studies, students and teachers of statistics, and anyone who needs to check the
equal-variance assumption before pooling data or running ANOVA.

• Works 100% offline • No sign-in • Light on storage