14 



STATION BULLETIN 355 



When yield and weight of drops are considered for the entire 

 period 1935 to 1941, there was a very high positive correlation with 

 Mcintosh, Delicious, Gravenstein, Starking, and Baldwin, the lowest 

 being with Spy (Table 10). It is certain that differences in correla- 

 tion coefficients within varieties for diff'erent years are due in part to 

 differences in seasonal effects and discrepancies with relation to har- 

 vest dates. It has been the practice at the horticultural farm to pick 

 Mcintosh over a longer period than other varieties, the lighter-yield- 

 ing trees and the higher-colored fruit usually being harvested first. 

 This was true in years of medium-to-heavy yield. Also, except in 

 1936 and 1941, the per cent drop increased as picking was delayed, 

 but not at the same rate in different years; nor was this difference 

 correlated with size of the crop. Climate and other factors must in- 

 fluence the amount of drop to account for these differences. 



Table 10. Correlation Between Yield and Weight of Drops 



Recently Welker and Wyndi" have emphasized that "groups 

 of biological data obtained under diff'erent conditions cannot be 

 pooled in an effort to increase the significance of statistical correla- 

 tions by increasing the number of pairs of variates." Also that "fac- 

 tors of unknown nature may falsely exaggerate or diminish the nu- 

 merical magnitude of the observed correlation coefficient." Further- 

 more, "these factors may even change the algebraic sigh of correla- 

 tions obtained from the pooled data." 



The writer has presented the preceding section on correlation 

 between yield and drops especially to bring out the fallacy of using 

 correlation data in interpreting any effect of yield on drops without 

 also having some knowledge of the other factors that may be aft'ect- 

 ing the relation but are not revealed by the correlation coefficient. 

 Such factors as soil moisture, nutrient supply, spray materials, weath- 

 er conditions etc., vary considerably from year to year. This causes 

 one to question the value of correlation coefficients obtained by using 

 pooled data obtained over a period of years. Similarly, in any given 

 year the picking date has a profound effect on the correlation between 

 drops and yield. Pooling the data from a number of trees in any 

 given year is also under suspicion since, even with a given variety, 

 fruit on different trees may not be at the same stage of maturity on 

 the same harvest date. Before using correlation data to substantiate 

 conclusions on the cause of any relationship one should, therefore, be 

 certain that all factors Avhich could possibly influence the relation are 



