141 



(compare what has been said in Remark II art. 14) and 

 do not show a well marked systematic change. In fact 

 we may say with considérable approach to truth that 

 the proportion is equal to the average value 



0.544 



throughout. 



The conclusion to which we are thus led would be, that 

 the différence in the distribution of the summer and 

 winter barometerheights can be explained by assuming 

 that they are governed by the same causes, which, 

 however, in summertime act with an intensity of only 

 about 54 '/o percent the intensity in winter time. 



It may contribute to a better understanding of the 

 meaning of proportional curves, if we compute the fre- 

 quency curve of the summer barometer readings theoreti- 

 cally from the winterreading. This computation offers no 

 difficulty provided we first dérive empirically tivo numbers 

 from a comparison of the summer and winter observations. 

 The first is the number / = 0.544 already fouud. This is 

 sufficient for the computation of the Zs by 



^^^ ''' = o:^5l4 • 



The further computation now becomes the inverse of that 

 foUowed before, when we derived the z' from the obser- 

 vations. From the z' (see tab. 9) we first obtain the z. 

 From thèse we then dérive the values of the scheme 

 and thèse finally yield the frequencies. 



In passing from the z's to the Zs we will want the 

 second of the necessary numbers. For as the z' are simply 

 the différences of the z, in order that the z' may be 



_r,-^-:r times greater, the z themselves mustbe _,_.. times 

 0.544 0.544 



greater. Besides, however, the z may ail be increased by 



the same amount A. For it is évident that such an increase 



