14 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 'J'J 



What is to be done? Certainly not to leave a known source of 

 error without attempting to remove it. We proceed as astronomers 

 do in correcting star observations for known defects. We separate 

 the results into groups with steadily mounting values of water vapor, 

 plot them, and determine the best corrections for the water vapor 

 effect. That was all we did or could do to correct such systematic 

 errors in the Mount Wilson data. 



This would be satisfactory if it were not for the sun's variation 

 at the same time. If we had 50 years of homogeneous observations 

 made at one station to discuss, solar variation could be neglected. 

 It would fall out in the means. But we cannot wait 50 years. 



Now comes the part that our critic objects to. We make only one 

 assumption. It is this : A series of observations taken with identical 

 water vapor, identical sky brightness, and all at one observatory, are 

 comparable without any corrections at all. Suppose we take all the 

 observations of one observatory and divide them into such groups, 

 each including only a very narrow range of humidity and sky bright- 

 ness. Each group of days indicates the solar variation in that group. 

 But there is no way to pass from one group to another, so long as we 

 have only one observatory. 



But arrange the values similarly for the other observatory. Again 

 we shall have the variation of the sun indicated strictly within each 

 group, but have no means to pass from group to group. But stay ! 

 The days comparable at one observatory fall in various groups at 

 the other. Thus, we find a great many independent determinations, 

 sometimes as many as 20, of each crossing-over factor from one 

 group to another. We take their mean indications, and so are able 

 at length to put all of the observations at each observatory on a 

 comparable footing. Then we compare all the days which are 

 common to both stations and we find that a small, uniform, constant 

 correction, which, of course, does not affect variability at all, is 

 needed to bring them to a common scale. 



In all this there is nothing that I can see to make Harqua Hala 

 variations dependent at all on Montezuma variations. After thus 

 getting all past observations to a comparable status, we can now 

 go back to eliminate solar variation from the original observations. 

 Having done this, we can go on with each station independently by 

 the usual method of grouping, already explained, so as to get a 

 separate formula for each station, by which all future observations 

 of that station are corrected. This also introduces no dependence 

 of one station on the other. 



