Cycles of Rainfall 19 



the ratio were say, 7 to 1, a greater ratio would not 

 occur by chance once in a thousand times. 1 



With these facts in mind, let us again examine Fig- 

 ure 3. It is clear that the principal periods needing 

 attention are those respectively of 8, 29, 33, 36 years. 

 In case of the 8 year cycle there can be very little 

 doubt as to the existence of a true periodicity approx- 

 imating 8 years in length. The ratio of the square of 

 its amplitude to the mean square amplitude of the 

 periodogram is 6.71 to 1. We may accordingly accept 

 with considerable confidence the existence of a natural 

 period of rainfall in the Ohio Valley approximating 

 8 years in length. 



The cycle of 33 years, inasmuch as the ratio of the 

 square of its amplitude to the mean square amplitude 

 of the periodogram is 3.27 to 1 is in all probability a 

 true cycle. The doubt that exists is due to the smallness 

 of the ratio and the few recurrences — only two 2 — 



1 Schuster: "The Periodogram of Magnetic Declination," pp. 124- 

 125. 



2 Those who deprecate the use of such meager data should con- 

 sider well the testimony of Lord Kelvin before the Meteorological 

 Committee of the Royal Society, 1876. 



Question 1710. "The sum which parliament will give for this 

 purpose being a limited sum, do you think that it would be well to 

 reduce the number of observations in order to have more money to 

 spend upon the reduction of observations? / think at all events until 

 one eleven years -period, the sun spot period, is completed, it would be 

 wrong to reduce the number of observations." 



Question 1735. "Supposing that you had one of these analyses 

 calculated for a period of 11 years, would each year's observations 

 and still more each period of 11 years observations, require to be 

 introduced into this analysis so that you would have an analysis of 

 22 years, and an analysis of 33 years, and so on from time to time, 



