TABLE OF CONTENTS In the preceding three chapters all the experiments involved a single factor and its effect on a measured variable. Several different designs were considered, but they all represented restrictions on the randomization in which interest was still centered on the effect of a single factor.
Suppose there are now two factors of interest to the experimenter: for example, the effect of both temperature and altitude on the current flow in an integrated circuit. One traditional method is to hold altitude constant and vary the temperature and then hold temperature constant and change the altitude, or, in general, hold all factors constant except one and take current flow readings for several levels of this one factor, then choose another factor to vary, holding all others constant, and so forth. To examine this type of experimentation, consider a very simple example in which temperature is to be set at 25 and 55 C only and altitudes of 0 K (K = 10,000 feet) and 3 K (30,000 feet) are to be used. If one factor is to be varied at a time, the altitude may be set for sea level or 0 K and the temperature varied from 25 C to 55 C. Suppose the current flow changed from 210 mA to 240 mA with this temperature increase. Now there is no way to assess whether this 30 mA increase is real or due to chance. Unless there is availabel some previous estimate of the error variability, the experiment must...
