Bergenlus et al Use of otolith morphology to indicate stock structure of Plectropomus leopardus 



505 



Discriminant analyses 



Two stepwise canonical discriminant analyses (CDAs) 

 were computed for the morphological variables ( Fourier 

 harmonics and shape variables combined) in response 

 to the significant cohort xcegion effect in the MANOVA 

 of the PC scores. First, separate CDAs were done for 

 each region using cohort as the separating variable 

 and second, separate CDAs were done for each cohort 

 by using region as the separating variable. The region- 

 specific CDA, where cohort was used as a separating 

 variable, showed a significant first discriminant func- 



tion (DF) for all four regions, accounting for 100% of 

 the variance in the data (Table 6). Some separation was 

 observed in the distribution of the DF I scores between 

 the two cohorts within each region, although the tem- 

 poral pattern was not consistent across regions (Fig. 

 4). Likewise, the cohort specific CDA using region as a 

 separating variable showed some separation between 

 the four regions in the first two significant DFs of 

 cohort 1995 (explaining 59.9% and 30.7% of the vari- 

 ance, respectively) and one significant function of cohort 

 1999 (explaining 100% of the variance. Table 6). The 

 means of the DF I scores in the Lizard Island and 



