Why It Matters: F Distribution and One-Way ANOVA

How can we analyze means if we have more than two populations to compare?

Often data is gathered across several groups and we wish to make inferences about the population means for those groups. Rather than conducting multiple hypothesis tests to compare two means, an Analysis of Variance (or ANOVA) is conducted. This requires a new statistic, known as the F statistic, and a new distribution, known as the F distribution. For example, based on these boxplots for the GPAs of a random sample of students at four colleges, do we think the mean GPAs are different at these colleges? Although we can see differences within the plots, an ANOVA would need to be done to completely answer this question.

Four parallel boxplots are shown. The boxplots are labeled from the bottom to the top as College A, College B, College C and College D. The boxplots for College A and B are similar, but College A has a slightly higher maximum (about 3.3) and third quartile (about 3.1) than College B. The distribution of GPAs of College C shows 50% of students with GPAs between about 2.4 and 2.6 and 50% of GPAs between about 2.6 and 4.0. The distribution of GPAs of College D has the highest median of all the colleges at about 3.2. The minimum and maximum GPA’s of College D are about 2.25 and 3.8, respectively.