RI 6232 An Introduction To Statistical Design Of Experiments In Metallurgical Research ? Summary

This Bureau of Mines report presents a discussion of the statistical approach to the design of experiments and indicates, by concrete examples, how and when statistical design is essential for obtaining meaningful data with a minimum number of experiments. Every research worker plans his work and designs his experiments in a manner which, based on his experience, seems most rational to him and which he believes will produce the information essential for resolving the problems. If the results obtained are open to multiple interpretations and are not of definitive reliability, the failure of the experiment design to produce the desired information may be due to several causes, These include faulty judgment in selecting the factors to be tested, ignoring the importance of interaction between factors, failure to adequately control test variables, and assuming that meaningful trends in the results can be identified without comparison to a predetermined experimental error. An experienced researcher probably will deduce from the nature of the data obtained that the design is faulty and may by trial and error ultimately arrive at the correct solution to the problem. Statistical methods of planning research minimize the chance of faulty experiment design, embody concrete procedures for testing the validity and reliability of the test data and for indicating mandatory changes in the design, and enable the researcher to obtain meaningful information with the least number of tests.