5.30 Randomization at work. To demonstrate how randomization reduces confounding, consider the following situation. A nutrition experimenter intends to compare the weight gain of prematurely born infants fed Diet A with those fed Diet B. To do this, she will feed each diet to 10 prematurely born infants whose parents have enrolled them in the study. She has available 10 baby girls and 10 baby boys. The researcher is concerned that baby boys may respond more favorably to the diets, so if all the baby boys were fed Diet A, the experiment would be biased in favor of Diet A.
(a) Label the infants 00, 01, . . . , 19. Use Table A to assign 10 infants to Diet A. Or, if you have access to statistical software, use it to assign 10 infants to Diet A. Do this four times, using different parts of the table (or different runs of your software), and write down the four groups assigned to Diet A.
(b) The infants labeled 10, 11, 12, 13, 14, 15, 16, 17, 18, and 19 are the 10 baby boys. How many of these infants were in each of the four Diet A groups that you generated? What was the average number of baby boys assigned to Diet A? What does this suggest about the effect of randomization on reducing confounding?