to determine what criteria were used to select specimens in 1977 , the 

 relationships were recalculated separately before combination. Ricker 

 (1975) noted that a functional geometric mean regression was more appro- 

 priate for these types of relationships than an ordinary predictive 

 regression which has been used to date. Therefore a geometric mean 

 regression was also calculated for the combined length-weight and 

 length-fecundity data. A comparison was then made among all the calculated 

 regressions by using three different lengths with each regression and 

 noting the percent difference between the 1977 relationships used most 

 frequently in previous reports and the others. The 1981 winter flounder 

 population survey data were also used with three length-fecundity regres- 

 sions to examine the percent difference in total and mean fecundity 

 produced with each. 



Small (3%) differences were found between the length-weight relation- 

 ship calculated for Millstone winter flounder between 1977 and 1982 

 (Table 21 ). Larger (-20-0%) differences were found between the weights 

 calculated from the arithmetic relationships for the Niantic River fish 



in 1977 and those from the 1982 logarithmic relationship. The combined 



lit 2 



Millstone and Niantic River data produced no change in r and coefficients 



of variation. Little differences were also found among the ordinary 



regressions and the geometric mean regression. Largest differences 



between the 1977 Millstone relationship used most frequently and the 



others occurred at the smallest (10 cm) and largest (40 cm) sizes 



examined. The combined data base using a geometric mean regression is 



probably the most correct one to use and produces a 11% increase in 



weight at 10 cm and a 7% decrease at 40 cm in comparison to the 1977 



Millstone standard; 



54 



