Brodziak and Macy: Growth of Loligo pealei 



223 



normal error structure could not be rejected for any 

 subset of the length-at-age data. Similarly, for all 10 

 subsets of weight-at-age data, the hypothesis of an 

 additive normal error structure was rejected at the 

 57c level, whereas the hypothesis of a multiplicative 

 lognormal error structure could not be rejected for 

 any subset. As a result, the multiplicative error struc- 

 ture was considered to be the best assumption for 

 modelling variability in size at age. 



Although the residuals from the curves estimated 

 with the multiplicative error structure conformed to 

 model assumptions, the parameters of these curves 

 were imprecisely determined. In particular, the case-I 

 curve was rejected for each subset of length-at-age 



and weight-at-age data because there was at least 

 one parameter that was not significantly different 

 from 0. Thus, the 4-parameter form of the Schnute 

 model had more parameters than necessary to char- 

 acterize L. pealei growth. 



For the length-at-age growth curves, most of the 

 case-II and case-Ill curves were rejected because 

 their parameters were not significant. As a result, 

 there were only 3 instances where comparison of RSS 

 was used to select the best curve. The best curve for 

 pooled-sex samples hatched during June-October 

 was the case-II curve which was selected over the 

 case-IV curve because the hypothesis a=0 was re- 

 jected (/"= 14.98 > 3.91=F 005 ( 1,142)). The best curve 



