W.A. Shewhart

Walter Shewhart was born in New Canton, Illinois on March 18, 1891 and died March 11, 1967 in Troy Hills, New Jersey. Shewhart, who taught and worked with W.E. Deming, was best known for the invention of the statistical control of quality, which u sed statistical methods to reach a state of control and judge when the state had been achieved.

Shewhart earned degrees from the University of Illinois and got his Ph.D. in physics from the University of California in 1917. After working as an engineer for Western Electric, he joined Bell Laboratories in 1925 where he spent the majority of h is career. Here, he used statistical tools to determine when corrective actions needed to be applied to processes in the lab. Throughout his career, Shewhart served as consultant for such bodies as the War Department, the United Nations, and the Indian government. He also taught at the Universities of Illinois and California, and was involved with various other universities such as Harvard, Rutgers and Princeton.

Shewhart's most important contribution to both statistics and industry was the development of the statistical control of quality. This idea incorporated the use of independent identically distributed random variables. The general principle behind this idea is that when a process is in a state of control and following any particular distribution with certain parameters, the purpose is to determine when the process departs from this control state so corrective action may be taken. Therefore, schem es must be developed that give a signal as to when the process has departed.

The Shewhart procedure, developed in 1924, addressed this matter. This procedure gave a signal when the process moved from the target mean m. Shewhart developed "action lines" at m ± k< FONT FACE="Symbol">s/Ö`n , where s is the standard deviation and k is a constant Shewhart found to be about 3. If the process departed from the target mean, a signal was given if the sample mean fell outside the action lines. This procedure was reliable enough to only have a probability of 1/500 of getting a Type I error. Shewhart displayed this idea of statistical control through the use of the control chart or the run-chart, which he proposed to his super iors on May 16, 1924.

One problem however in Shewhart's procedure was that it did not find the magnitude of change in the process, and it was unable to quickly find large changes within small samples. Knowing the magnitude of change would allow one to adjust the proced ure by the magnitude found. These problems were addressed by statisticians such as Dudding, Jennett, and Grant in the 1940's and 50's.

Shewhart made other contributions to statistical methods as well. He talked about the need for operational definitions and specifications in reportings on research. Characteristics of data (flat, blue, male, female, etc.) can not be communicated properly unless it is in statistical terms. Characteristics have no true value on their own. Shewhart also believed that in reporting research results, the data given should preserve all the evidence. Statistical parameters such as average and variance should only be used if they lead back to the same results. Also, being skilled in physics, Shewhart stressed his idea that physical laws only make perfect sense in statistical contexts. He claimed too many constants are used where as in real life situa tions, these constants rarely hold. Only through the use of statistics can one truly get accurate results of these varying physical laws.

Shewhart's contributions to the world of statistics and also to industry have proved very significant. His influence on statisticians such as W.E. Deming resulted in process improvements and higher quality in industry that led to the Japanese econ omic boom of the twentieth century. His influence will not soon be forgotten.