The F-test for variances is used to decide whether two populations appear to have the same spread. It compares two sample variances by forming a ratio, so it is useful when variability itself is the question. This matters in science, engineering, manufacturing, and statistics because two processes can have the same average but very different consistency.
The test helps answer whether an observed difference in sample variance is large enough to be unlikely from random sampling alone.
The test statistic is F = s1^2 / s2^2, where s1^2 and s2^2 are sample variances from two independent samples. Under the null hypothesis that the population variances are equal, this ratio follows an F-distribution with degrees of freedom df1 = n1 - 1 and df2 = n2 - 1. Because variances are always nonnegative, F-values are always positive and the F-distribution is right-skewed.
A worked example usually places the larger sample variance in the numerator, computes F, then compares it with a critical value or p-value.
Understanding Statistics: The F-Test for Variances
Variance measures the typical squared distance of values from their sample mean. Squaring is important because positive and negative differences would otherwise cancel. A few values far from the mean can contribute a great deal to the variance.
This makes a variance comparison different from comparing averages. Two machines may fill bottles to the same mean volume, yet one machine may produce many more bottles that are noticeably underfilled or overfilled.
In that setting, the more variable machine can create waste even when its average output looks correct. Students should connect spread to reliability, precision, and quality control rather than treating it as just another number from a calculator.
The F distribution describes what can happen to a variance ratio when the two underlying population variances really are equal. Sampling causes the two sample variances to differ naturally. Small samples show especially large changes from one sample to another, so they need stronger evidence before suggesting a real difference in spread.
The degrees of freedom describe how much independent information each sample provides for estimating its variance. Larger degrees of freedom make the F distribution less spread out.
Its shape depends on both samples, so changing either sample size changes the critical values. Tables and software must use the numerator degrees of freedom in the correct position and the denominator degrees of freedom in the correct position.
A two-sided variance test checks for a difference in either direction. Placing the larger observed variance on top is convenient because attention can focus on the upper tail of the distribution. The p-value is the probability of getting a ratio at least as extreme as the observed ratio if equal population variances were true.
A small p-value is evidence against equal variances, but it does not measure the practical importance of the difference. A very large study can detect a small difference that has little real effect.
Report the sample variances, sample sizes, ratio, p-value, and the context of the measurements. This gives readers enough information to judge statistical evidence alongside practical consequences.
The normality assumption deserves careful attention. The F-test is sensitive to outliers and strongly skewed data because variance itself is sensitive to extreme observations. One recording error, an unusual batch, or a rare event can make a sample variance much larger.
Before using the test, inspect a histogram, box plot, or ordered list of values. Check that observations were collected independently. Measurements from the same person over time, paired products, or repeated readings from one device are not independent samples.
When data are clearly non-normal or contain outliers, methods such as Levene's test or Brown-Forsythe test are often safer choices. In later statistics courses, variance comparisons matter when choosing versions of mean comparison procedures and when checking assumptions for regression and analysis of variance.
Key Facts
- Test statistic: F = s1^2 / s2^2.
- Null hypothesis for equal variances: H0: sigma1^2 = sigma2^2.
- Degrees of freedom: df1 = n1 - 1 and df2 = n2 - 1.
- If the larger sample variance is placed in the numerator, then F >= 1.
- The F-test assumes independent random samples from populations that are approximately normal.
- For a two-tailed test at significance level alpha, reject H0 if F is too large or too small compared with F critical values.
Vocabulary
- Variance
- Variance measures the average squared distance of data values from their mean.
- F-statistic
- The F-statistic is the ratio of two sample variances used to test whether population variances are equal.
- F-distribution
- The F-distribution is a right-skewed probability distribution used for ratios of variances.
- Degrees of freedom
- Degrees of freedom describe how many independent pieces of information are available for estimating a statistic.
- Significance level
- The significance level alpha is the chosen probability of rejecting a true null hypothesis.
Common Mistakes to Avoid
- Using standard deviations directly in F = s1^2 / s2^2 is wrong because the F-test compares variances, not standard deviations. Square each sample standard deviation before forming the ratio.
- Forgetting the degrees of freedom order is wrong because df1 must match the numerator variance and df2 must match the denominator variance. Swapping them changes the critical value and p-value.
- Applying the F-test to strongly nonnormal data is risky because the test is very sensitive to departures from normality. For skewed data or outliers, consider a more robust method such as Levene's test.
- Treating a large F-value as proof of different means is wrong because the F-test for variances tests spread, not center. A separate test is needed to compare means.
Practice Questions
- 1 Sample A has n1 = 12 and s1^2 = 45. Sample B has n2 = 10 and s2^2 = 18. Compute F using the larger variance in the numerator and state df1 and df2.
- 2 Two machines produce parts with sample standard deviations 3.2 mm and 2.0 mm from samples of sizes 16 and 11. Compute the F-statistic using the larger variance in the numerator and give the corresponding degrees of freedom.
- 3 A data set has several extreme outliers and a strongly right-skewed distribution. Explain why a standard F-test for equal variances may be unreliable and what assumption is being threatened.