78 



BELL SYSTEM TECHNICAL JOURNAL 



(column 4) bjiscd upon the assumption of the second api>ro.\iniation is 

 also a very close fit to the obser\ed frequencies, the ])r<jbability of fit 

 being in this case .69. As a result of these calculations shall we con- 

 clude that the distribution is composed of two normal components as 

 indicated in F'ig. 10, or shall we conclude that the distribution is homo- 

 geneous? In other words, do the skulls belong to two or to only 



Fii;. 10 



one race? Tlie measure giwn 1)\ the probaliilily of fu is, of course, 

 in favor of the first alternative. It is hii^hly ])rol)al)le, howe\-er, that 

 if we had been given the obser\'ed distributioTi without any discussion 

 of what it meant we would ha\e decided that it probabh- was con- 

 sistent with the assumption of the random s\-stem of causes such as 

 might underlie the second approximation. 



In other worils, if we had been gi\'en nuTely the above set ol skull 

 measurements, it is reasonable to suppose we might have concluded 

 that the distribution was homogeneous. However, when our judgment 

 is colored l)y the facts which cannot be presented in the arraN- of 

 observed frequencies we must conclude that it is highly probable 

 that the obserxi'd d.ita lia\e arisen from a non-homogcncoiis i)op- 

 ulation. 



Statistical methods alone do not answer all of llic (|urstioiis that 

 aie raised in this problem nor do the>' answer ilieni in nuun others. 

 There is almost always room for judgment to enter. 



Thus, analyzing a group of measurements of some charaiteristic 

 of a large number of transmitters, it often becomes necessary to 

 determine whether or not they can be subdi\'ided into normal coin- 



