ECONOMIC QUALITY CONTROL OF PRODUCT 381 



of material used in aeroplane propellers during the war. The wide 

 variability is certainly striking. The smooth solid curve is an approx- 

 imation to the distribution function for this particular property repre- 

 senting at least approximately a state of control. The importance of 

 going from the sample to the smooth distribution is at once apparent 

 and in this case a comparatively small amount of refinement in statisti- 

 cal machinery is required. 



Suppose, however, that instead of more than a thousand measure- 

 ments we had only a very small number, such as is so often the case in 

 engineering work. Our estimation of the variability of the distribution 

 function, representing the state of control, upon the basis of the inform- 

 ation given by the sample would necessarily be quite different from that 

 ordinarily used by engineers (see Fig. 10). This is true even though 



i.Sr 



Z 1.6 



o 



1- 

 < 



> 



LU 

 Q 



gl.4 

 < 



Z 

 < 



I- 

 (/) 



ti. 

 O 1.2 



UJ 



I- 

 < 



\- 



LJ 



LOh 



INEW 



OLD 



i- 1 ' 1 1 1 ^ L_^ 1 1 \ I I I 1 I I I I 



3 5 10 20 50 100 



n 



Fig. lO^Correction factors made possible by modern statistical theorj' are often 



large. — ^Typical Illustration. 



we make the same kind of assumption to begin with as engineers have 

 been accustomed to do in the past. This we may take as a typical 

 example of the fact that the production engineer finds it to his advan- 

 tage to keep abreast of the developments in statistical theory. Here 

 we use new in the sense that much of modern statistical machinery is 

 new to most engineers. 



