Abstract
Large masses of data in industrial experimentation are meaningless unless they can be reduced to a form from which it is possible to readily estimate the effect of existing conditions on the process or product. In the past, multiple-regression analysis and statistical quality control have been used in this connection with limited success. It is now possible to obtain more complete results through the use of a method developed by mathematical statisticians. This method is outlined and discussed in detail. It is shown how use of this technique enables an investigator to use each piece of data for making each conclusion and enables him to examine the behavior of one variable at more than one level of each of the other variables in the study. 2.1.1
