CORRECTIOiX or DATA fOR M EAST REM I-.XT 



19 



Summary 



We are now in a position to summarize the practical routine to be 

 followed in finding the most jirobahle distribution friX) of quality 

 when the observed distribution is given. 



To find /r(X), we must first know the law of error //i(.v). We 

 must show this to be normal and find the standard deviation <^e 



EFFICIENCY 



Fig. 6 — Correction which must be applied to the observed distribution of transmitters 

 Fig. 1, because of the existence of errors of measurement 



by multiple tests on a single unit. The error made in determining 

 the standard deviation <^e from 7i observations is y— . Hence the pre- 

 cision we attain in finding fr{X) depends upon the number of obser- 

 vations n made in finding tr^-. 



Having found (^e to the required degree of precision, we must next 

 discover whether or not the observed distribution Jt{X) is either 

 normal or the second approximation. Standard statistical methods 

 can be used for this purpose. 



If the/o(A") is normal, we then know that 



MX) 



X' 



\/2T{as~ald 



e 2(<TZ-aE)< 



and, if / {X) is second approximation, we know that friX) is given 

 by Equation (6), where or and kr can be found with the aid of Equa- 



