Austin et al.: Morphometnc separation of annual cohorts within mid-Atlantic Pomatomus saltatnx 



413 



which more clearly shows Y^ is an estimate of the 

 average Y^ for an individual of fork length X^. 



Following transformation, each variable was re- 

 gressed against fork length (FL). The slope of each 

 transformed variable on FL was zero or insignificant 

 in all cases; therefore effects of allometry were dis- 

 regarded. The results were plotted for visual inspec- 

 tion of outliers which were removed before subse- 

 quent analyses if they were outside the range of bio- 

 logical possibility, and thus suggested measurement 

 error. Consequently, two subgroups of data were de- 

 veloped. The first included bluefish between 200 and 

 400 mm fork length (yearlings), to remove young-of- 

 the-year from the data which are in the stage of 

 growth most likely to show allometry and to dupli- 



cate the size range used by Wilk and Walford in 1964 

 (Wilk, 19771. The second group included bluefish 

 greater than 400 mm fork length. 



The linear discriminant function used here is of 

 the form 



D = BiXj + B.2X., + B3X3 + B„X„+C 



and is similar to a multiple linear regression 



where D = the discriminant function that char- 

 acterizes each reference group; 

 X 's - theindependent variables (individual 

 measurements) selected at in a step- 

 wise fashion; and 



