ABSTRACT
WH E N W E C O M P A R E more than two groups in an ANOVA a signifi-cant F value does not indicate where the effect lies, simply that there is an effect between the conditions somewhere. A researcher compared four groups of children (6, 8, 10 and 12 year olds) on a test of social skills. She found a significant F value and concluded that the scores from the four conditions were not drawn from the same distribution. But this conclusion does not really provide the researcher with the information about which ages show the significant differences. Let us assume that the means were respectively 10, 12, 18 and 23 on the test (out of 50). Given that there is a significance variance ratio it seems likely that the scores of the 6 year olds differ significantly from those of the 12 year olds as this comparison provides the largest difference in means. Is the difference between the 6 and 8 year olds or 8 year olds and 10 year olds significant? And what about the smallest difference, between the 6 and 8 year olds? The data needs to be inspected further to find the source of the significant F value.
