26 MAINE AGRICULTURAL EXPERIMENT STATION. I915. 



minus selections but with the plus selections the opposite result 

 is obtained. 



In view of the possible significance of these figures and those 

 exhibited in tables 3 to 6 it is worth while to^ attempt an analysis 

 of these data by still other methods. 



Indices of Sele:ction. 



Attention has been called to the fact that in the method used 

 in the preceding pages two factors have been left out of account. 

 These are ( i ) the possible difference in the variability of differ- 

 ent pure lines, and (2) the size of the deviation of the selected 

 plants. That is, all plus deviations of the selected plants have 

 been given equal weight although some plants were 20 grams 

 above the mean and others only a fraction of a gram. If there 

 is any real influence of the selection, a plant 20 grams above its 

 mean should have more effect on the next generation than a 

 plant only one-half gram above its mean. In view of the fact 

 that the results obtained by the preceding method are somewhat 

 doubtful in meaning it will be well to take these factors into 

 consideration. 



The difference in the variability of the different pure hnes 

 could best be expressed by their standard deviations or coeffi- 

 cients of variation. It would then be possible to divide the dev- 

 iation of each selected plant or of each row by the standard 

 deviation of the corresponding pure line. This would express 

 the variability of the plant (or row) in terms of the varia- 

 bility of the pure line. However, in the majority of the pure 

 limes used in this work the number oif rows is too small to obtain 

 a reliable standard deviation. 



Very probably the chief factor in increasing or decreasing the 

 standard deviation of a given line is the size of the mean. The 

 greater the size of the mean the greater the chance for absolute 

 variation. Consequently if each deviation is expressed as a per- 

 centage of its mean it is probable that the chief elcrment in 

 the difference in variability will have been taken into account. 



We have, therefore, gone through the tablesi (cf. table 2) for 

 each pure line and expressed the deviation of each plant and 

 each row as a per cent, of the mean of that pure line for the 

 given year. 



