ANALYSIS OF TERRESTRIAL RECORDS 



93 



have also prepared a scatter diagram of relation between successive depar- 

 tures from a Hanned mean of random values in figure 40b, which gives a 

 circular cloud of dots and therefore is without indication of negative correla- 

 tion in successive values. 



The solution of the two-year cycle with its difficulties of distribution 

 about the earth has a reasonable chance of being successful; we have now 

 an idea of its general action, and by prediction on the basis of short cycles 

 and then applying the probable residual derived from the inclined axis of 

 figure 40a, we reach the most probable value at this time available. 



40 



!20 



•oo 



•I • 



I c20 

 t-'2 



/ 



/ ° 



\> „ 



40 — 



40 



— ^- 



i : 



\ 



\ * 



\ 



^T 



\ 



y .\ 



v 



.«••. 





3^ 



a ft 



20 20 



Minus Plus 



The year itself 

 CL 



40 40 20 20 



Minus Plus 



The year itself 

 ft 



Fig. 40 — Successive year conservation in departures from a smoothed curve, 

 Fort Defiance group (a) in natural data; (b) in random data. Same 

 scale as fig. 39. 



CYCLES IN 42 WESTERN GROUPS 



The groups here listed are indicated by the letters used in the former re- 

 port, Volume II, Climatic Cycles and Tree Growth, Chapter VII, to which 

 reference is made for details of location and collection of specimens and re- 

 duction of data. 



The assignment below of exact values in cycle length has seemed advisable 

 in order to show the figures on which amplitudes have been computed. Re- 

 membering that cycle lengths are subject to errors of 1 to 3 per cent, the 

 reader will be able to find identity in many of these that differ less than this 

 percentage. The cycle lengths given here are used to form the periodograms, 

 figure 41. In these tables the underlines may be taken as nearly equivalent 

 to weights. The rows, also sometimes called "sets," are the number of 

 repetitions of the cycle length. The second and last columns refer to the 

 summation curves in the appendix (see p. 155). Some of the results here 

 presented will be slightly modified after subtracting dominating interfering 

 cycles. It should be recognized that this is the first attempt to group a large 

 number of discontinuous periods together. We expect that more experience 

 will show better how to do it. Therefore we must regard this as a series of 



