The Flight of Nol^turnal Lepidoptera 51 



In working with the figures presented in Table II, the data for 

 the seven nights marked with an asterisk (*) were omitted, as showing 

 the effects of precipitation, leaving a total of eighty observations, on 

 which all computations are based. 



A comparison of Curve I. Plate \^I. with the weather curves of the 

 same plate makes it quite evident that weather factors have a consider- 

 able influence on the size of the catch, but the effects are somewhat 

 obscured by the variations in the curve caused by the irregular emer- 

 gence of the various species. In order to show the fluctuations due to 

 weather alone. Plate VII was drawn, on which the first curve repre- 

 sents the per cent of "normal" catch (Column 6, Table II). I'he re- 

 maining weather curves are from the same data given in Table II. A 

 study of this graph brings out the following tendencies. 



1. The fluctuations in catches can not be closely correlated with 

 any one of the w^eather curves, but the catch apparently tends to increase 

 with a rise in temperature, and decreases as either humidity or pressure 

 deviates very far in either direction from the optimum indicated by 

 parallel lines. 



2. The relative efifects of the various factors are greater, the far- 

 ther the condition departs from the optimum. 



3. The largest catches were made when all factors were at or 

 near their respective optima. 



In order to show these effects more clearly, charts were made by 

 plotting the percentage of normal catch against temperature, humidity, 

 and pressure in turn. These charts are shown on Plate \\l. numbered 

 V, VI. and VII. The tendencies mentioned show up (juite clearly, 

 especially the eft'ect of the optimum humidity condition. 



To bring out the relative values of the various factors, recourse 

 was had to statistical methods, and the method of partial correlation, 

 as developed by Yule (6) and as applied to crop yields and weather 

 factors by Blair (7. 8) was adopted. Without attempting to deal with 

 the mathematics of the method, it may be said that the correlation 

 coefficient is a pure number, mathematically obtained, which indicates 

 the tendency of one factor to vary with or in opposite direction to 

 another. If the factors are closely related, a value approaching unity 

 is obtained, and if totally unrelated, a value near to zero. The sign of 

 the coefficient indicates the character of the correlation, a plus ( + ) 

 sign meaning that an increase in one factor is related to an increase in 

 the other, and a minus (— ) sign that an increase in one factor is re- 

 lated to a decrease in the other. The notation employed is to designate 

 the coefficient as "r," with subletters or numbers indicating the factors 



