Test whether two groups have the same variance. Paste two data columns, get the F-statistic, p-value, variance ratio CI, and guidance on whether to use a pooled or Welch t-test.
Paste your two data columns and press Calculate to run the F-test for equal variances.
Watch: 1-Sample, Paired t-tests and Tests for Equal Variances
Two machines make the same part. Their averages look similar, but is one more consistent than the other? Enter the lab to compare their variation.
Paste measurements for two independent groups (one per line). They don't need the same size, but each needs at least 2 values.
The F-statistic is simply the bigger sample's spread divided by the smaller sample's spread. A large F means the two groups vary by very different amounts.
The p-value tells you whether the difference in spread is real or just chance. The result also recommends whether to use a standard or Welch t-test when you compare the two group averages next.
Use the calculator above to compare the spread of two groups and test whether the difference in variance is real or due to sampling chance. The F-test is essential when you want to choose between a pooled and a Welch t-test, and when variability itself is the quality characteristic you are trying to reduce.
The F-test for equal variances (also called the variance-ratio test) tests the null hypothesis that two populations have the same variance. You compute the ratio of the two sample variances: F = s₁²/s₂². If the variances are equal, F should be close to 1. A large F (or, equivalently, a small p-value) gives evidence that the variances genuinely differ — one group is more variable than the other.
The F-test compares how spread out two sets of measurements are. It calculates the variance (a measure of spread) for each sample, then divides the larger variance by the smaller one to get the F-ratio. If both machines were equally consistent that ratio would sit near 1; the further it climbs above 1, the more their consistency differs. The p-value tells you whether that gap is real or just sampling luck, and the confidence interval gives a plausible range for how many times more variable one is than the other.
Two CNC machines produce the same part. Machine A has a standard deviation of 0.8mm across 20 parts. Machine B has 1.4mm. Machine B looks more variable — but is the difference real with only 20 parts each?
The F-test gives p = 0.04, confirming Machine B is genuinely more variable, not just unlucky in this sample.
What to do with this: Book a maintenance check on Machine B. More variability usually means worn tooling, loose fixtures, or inconsistent feed rates — fix it before out-of-spec parts ship.
In manufacturing and quality, variability is often the real enemy — a process that is consistently wrong can be adjusted, but a variable one produces unpredictable output. The F-test quantifies whether one process is significantly more variable than another, directing improvement effort where variation is greatest.
Use the F-test before a 2-sample t-test to decide whether to pool variances. Also use it directly in the Analyse phase when the question is about consistency rather than mean levels — e.g. comparing two machines, two suppliers, or two operators for variability rather than average output.
In Six Sigma, the F-test helps identify which sources introduce more variation — a direct input to the Analyse phase. Combined with a capability study (Cp/Cpk), it shows not just whether a process is centred on target but whether its spread is acceptable and consistent across groups.
If F is significant, the variances differ: use Welch's t-test (not pooled) when comparing means, and investigate the operational cause of the higher variability. If F is not significant, the equal-variance assumption holds: use the pooled t-test. In either case, plot the data (box plot, histogram) alongside the test results to communicate the finding visually to stakeholders.
Combine the F-test with box plots, Cp/Cpk, and DMAIC structure for a complete variability-reduction workflow.
The F-test (Levene or variance-ratio test) tests whether two populations have the same variance — i.e. the same degree of spread around their means. A significant result means the variability differs between groups, which matters for choosing the right t-test (pooled vs Welch) and for understanding process consistency. In quality improvement, higher variance typically means less consistent output.
The F-statistic is the ratio of the two sample variances: F = s₁²/s₂², where s₁² is the larger of the two variances. By convention the larger is placed in the numerator so F ≥ 1. F follows an F-distribution with (n₁−1) and (n₂−1) degrees of freedom. A large F means the variances are very different; F near 1 means they are similar.
The 2-sample t-test assumes equal variances (pooled t-test) or allows for unequal variances (Welch's t-test). Running the F-test first tells you which t-test to use: if F is not significant, use the pooled t-test; if F is significant, use Welch's t-test which does not assume equal variances and is more robust.
The F-test assumes both samples come from normally distributed populations. It is quite sensitive to this assumption — more so than the t-test — so it is important to check normality first. If normality is doubtful, Levene's test (which is more robust to non-normality) is preferred. The samples must also be independent of each other.
Knowing which group is more variable is the start. Reducing that variability requires structured investigation and improvement tools.
More practical calculators for quality, statistics, and improvement.
Want to apply the F-test and full hypothesis testing toolkit inside a structured improvement project? The Black Belt course covers advanced statistics, hypothesis testing, and the full analytical toolkit.
View Black Belt →