In: Statistics and Probability
When we do find a significant ANOVA, we do not know where the difference lies. It could be that only two groups are significantly different, or that all three groups are significantly different. In your response you mention that one problem with t-tests is that they only compare two groups at a time. This creates another problem with conducting multiple t-tests, which is that it increases our chances of committing a type I error. In what way do post-hoc tests attempt to reduce this problem with conducting multiple t-tests?