143 



hâve to borrow from the observations. I find that every 



thing is best représentée! if we take 



{d) A - 0.016. 



With this value formula (c) furnishses ail the z^. 



Thèse being obtained we find the numbers in the 

 column "Scheme" by such a table as tab 11. In the présent 

 case it is necessary to use 4 décimais and I therefore 

 made use of the table at the end of the 2nd paper. — Having 

 got the scheme, we get the frequency curve by taking the 

 différences between consécutive values. 



The results obtained in this way are as shown by 

 table 3, third column. The second shows the observed 

 frequencies. The agreememt is surprising. 



Remark 1. If by such agreement we feel convinced that we 

 hâve really to do with proportional curves and if — by some 

 independent means — we could be sure that the propor- 

 tionality were of the first kind, then the curious resuit would 

 foUow, that we could détermine the undisturbed barometer- 

 height at den Helder, that is the reading the barometer 

 would show in the absence of any perturbing factors. 

 We must however refer to the 2nd paper for this matter. 



Remark 2, AU normal curves may be considered as 

 proportional curves. 



17. Médians and quartiles* 



If we call 

 ;co.25 the x for which corresp. value in scheme is 0.25 

 y- .. „ „ „ „ „ „ „ „ 0.50 



.X-0.75 „ „ „ „ „ „ „ .. .. 0.75 



then, according to difinition of the scheme 

 one fourth of ail the x lie below jco.25 



,, ,, ,, between xo.25 and y. 



„ „ ,, ,. „ ,. y- ,, .X-0.75 



and according to usage 



called the médian; 



xq.js is called the Ut quartile; 



.V0.75 — !■>■ ,, .. ., 2nd 

 Recueil des trav. bot. Néerl. Vol. XIII. 1916. 10 



( 'J- is 



