Home » Additional Features, Statisticians in History

Walter A. Shewhart, 1891–1967

1 September 2009 8,674 views No Comment
In 1967, W. Edwards Deming wrote a personal biography on Walter Shewhart for The American Statistician. The following is taken from that article.


shewhart

“It is his knowledge and use of the theory of probability that distinguishes the statistician from the expert in chemistry, agriculture, bacteriology, medicine, production, consumer research, engineering, or anything else.”


I write as one outside the Bell System who had the privilege of working intimately with [Walter] Shewhart over a period of years. This could happen only because he was always glad to help anyone. Actually, he never thought of himself as helping anyone; he was simply glad to talk and absorb thoughts from anyone [who] was genuinely struggling to improve his understanding of the statistical method—interchanging ideas was his way to put it. And, to Shewhart, it was the statistical method in the singular, not in the plural. Statistical methods are necessary, but they are the tools and passwords by which the statistician works and communicates in applying the statistical method.

It was Shewhart who emphasized the theory of probability as the tool of the statistician. It is his knowledge and use of the theory of probability that distinguishes the statistician from the expert in chemistry, agriculture, bacteriology, medicine, production, consumer research, engineering, or anything else. Otherwise, the statistician would be merely another chemist, another agricultural scientist, or something else.

Quality control meant to him [the] use of statistical methods all the way from raw material to consumer and back again—through redesign of product, reworking of specifications of raw materials—in a continuous cycle as results come in from consumer research and from other tests.

He was quick to see that quality must mean not necessarily high quality, but dependable and economic quality, which in turn meant quality suited to the purpose. But what quality is suited to the purpose? Statistical methods for discovery of what product is needed, what quality is needed, and for learning how a product performs in service and in the laboratory are thus necessary ingredients of the statistical control of quality.

The world knows him for the Shewhart control charts, and the world lives better for them. They are, however, only one of his statistical contributions. He leaves a rich legacy that will take years to absorb. For example, his rules 1 and 2 on the presentation of data:

Rule 1. Original data should be presented in a way that will preserve the evidence in the original data for all the predictions assumed to be useful.

Rule 2. Any summary of a distribution of numbers should not give an objective degree of belief in any one of the inferences or predictions to be made there, for that would cause human action significantly different from what this action would be if the original distribution had been taken as a basis for evidence.

Then, there is his Criterion of Meaning:

Every sentence, in order to have definite scientific meaning, must be practically or at least theoretically verifiable as either true or false upon the basis of experimental measurements, either practically or theoretically obtainable by carrying out a definite and previously specified operation in the future. The meaning of such a sentence is the method of its verification.

The above rules and criterion of meaning were, to him, a necessary ingredient of industrial research for the reason that, as he stated, industrial research is more exacting then pure science. His faith in the power of the statistical method in all human inquiry was unshakable.

He acknowledged an everlasting debt to C. I. Lewis’ Mind and the World Order, which he recommended to me. I had the usual difficulty with it, and I recall saying to Shewhart at the end of the seventh reading that so far it had meant nothing to me. “Stay with it,” he said. “I read it 14 times before it began to mean anything.” I wonder how he came upon it in the first place, and how he knew how important it was that he should pursue it.

Although operational definitions, his criterion of meaning, and his rules 1 and 2 for the presentation of data have been known to scientists for several generations, no one to my knowledge has stated them so well as Shewhart. One sees in them Lewis in the background.

Hypothesis Is Necessary

Some knowledge must be a priori, even if shown later by observation to be untenable. Without theory (hypothesis), data are meaningless or nonexistent. There is thus no true value of anything; true value is undefinable [sic] operationally. There are, however, numerical values that people can use with confidence if they understand their meaning (for the tensile strength of a batch of wire, for example, or for the proportion of the labor force unemployed last month).

There was to Shewhart no such thing as a random sample. There was and is, however, such a thing as a sample selected by a random operation. There may be a concept of randomness, but one cannot communicate it. What one can communicate is an operational definition of a random operation (for example, proper use of random numbers). Likewise, one can only define yellow, green, tired, unemployed, [or] one inch in terms of an operation. The particular operation will vary with the needs of the subject matter.

There is accordingly no such thing as factual information, distinguished from judgments. Physical measurements are no exception. There are no facts, except as man makes them. Man gets marks on a piece of paper in response to a stimulus. Such marks on paper and tabulations made from them are useful only if the method of investigation is suited to the purpose.

1 Star2 Stars3 Stars4 Stars5 Stars (1 votes, average: 5.00 out of 5)
Loading...

Comments are closed.