Multiple Comparison Techniques
In analyzing a linear model we can examine as many single degree of freedom hypotheses as we want. If we test all of these hypotheses at, say, the 0.05 level, then the (weak) experimentwise error rate (the probability of rejecting at least one of these hypotheses when all are true) will be greater than 0.05. Multiple comparison techniques are methods of performing the tests so that if all the hypotheses are true, then the probability of rejecting any of the hypotheses is no greater than some specified value, i.e., the experimentwise error rate is controlled.
Unable to display preview. Download preview PDF.