REPORT OF THE SECRETARY 109 



separate points. The area it includes over the line of zero depar- 

 tures is crosshatchcd. Obviously the Dia^^ram F has a preponder- 

 ance of crosshatched area below the zero line, and the Diafijram G 

 above. Yet in April, and from June to mid-November, these aspects 

 tend to reverse themselves in each diagram. To bring out this char- 

 acteristic more plainly, I give in each diagram curves made by 

 taking 5-case consecutive moans. That is, the mean is taken of cases 

 1 to 5, 2 to 7, 3 to 8, and so on. As it seemed clear that such a curve 

 in Diagram G is nearly the reverse of that in Diagram F, I have 

 inverted the 5-day mean curve of G as the dotted curve in F. The 

 correlation coefficient between the full and dotted curves in F is 

 50 ±5 per cent. 



Results are found for Yuma and AVilliston similar to those pre- 

 sented for AVashington. Though the types of elfect do not occur in 

 identical months at the three stations, the magnitudes and tendencies 

 are much the same. 



