12.41 Use computer software to work Exercise 12.37. Do’s and Don’ts Do’s 1. Whenever possible, randomize the assignment of treatments in the completely randomized design. In other designs, randomize the assignments of treatment within the restraints of the design. 2. When numerous comparisons must be made, use a multiple comparisons method to be able to make a confidence statement about the whole set of confidence intervals. 3. Consider applying the analysis of covariance when an important extraneous variable cannot be held constant. Don’ts 1. Don’t routinely accept the analysis of variance presented in a computer output. Instead, inspect the residuals for outliers and moderate to severe lack of normality. A normal-scores plot is useful if there are 20 or more residuals. It may suggest a transformation.