158 
Fishery Bulletin 11 7(3) 
Table 3 
Regression coefficients from analysis of the relationships of fecundity and inde¬ 
pendent variables of size and age for rainbow smelt (Osmerus mordax) collected 
in marine waters off Massachusetts during 2004-2007. Variables were log- 
transformed for all regressions. Standard errors of the mean (SEs) are given for 
intercept and slope values. All values for the E-statistic are significant (P<0.001). 
f? 2 =coefficient of multiple determination; $ 2 y . x =residual mean square. 
Variable 
n 
R 2 
q2 
° yx 
F 
Intercept (SE) 
Slope (SE) 
Length 
110 
0.891 
0.040 
884.9 
-6.404 (0.549) 
3.166(0.106) 
Total weight 
110 
0.899 
0.037 
961.1 
6.765 (0.104) 
0.927 (0.030) 
Ovarian weight 
110 
0.739 
0.096 
305.7 
9.411 (0.042) 
0.518(0.030) 
Age 
107 
0.524 
0.174 
115.4 
7.721 (0.209) 
2.062(0.192) 
11.0 
5 105 
c 10.0 
D 
s?. 
£ 9-5 
• 5* 
o 9.0 
iffp 
• 
8.5 
• 
• 
• 
• 
4.8 5.0 5.2 5.4 
3 4 
log (Length) 
log (Total weight) 
log (Fecundity) 
CD CO O O —*• 
o cn b cn o 
1 
A 
i 
1 ! ! 1 
:• r u 
9 # 
8.5 
• 
• 
• 
• 
0.0 0.5 
1.0 
-10 12 3 
log (Age) 
log (Ovarian weight) 
Figure 4 
Regressions of the relationships between fecundity and the 
size and age of rainbow smelt (Osmerus mordax) sampled 
in marine waters off Massachusetts during a trawl sur¬ 
vey in 2004-2007. See Table 2 for summary statistics from 
fecundity analysis and Table 3 for regression coefficients. 
The vertical clusters of points in the log (Age) graph repre¬ 
sent ages 1-4 from left to right, respectively. The dashed 
lines indicate the fitted regression model values. Lengths 
are given as total lengths in millimeters, and weights are 
total weights given in grams. 
net sampling at the location in the Fore River had the 
highest total catch, frequency of occurrence, and catch 
per unit of effort (number of rainbow smelt per overnight 
haul) among the 4 rivers. The results that follow include 
comparisons of size and age data for rainbow smelt caught 
during the trawl survey and fyke net sampling. 
Length and weight correction A random sample of rainbow 
smelt was collected with the fyke net in the Fore River to 
measure length and weight when fish were fresh (<12 h 
after collection) and again after they had been frozen for 
3 months. Regressions were generated for both TL and TW. 
No differences were found in the regression slopes between 
sexes through analysis of covariance. Therefore, the rela¬ 
tionship between data for fresh and frozen samples that 
was used to correct the TL of trawl samples was determined 
with this equation: TL Fresh =-1.88893+0.99075TL Frozen (n=59, 
jR 2 =0.985, PcO.OOl). The following equation was used to 
determine the relationship between data for fresh and frozen 
samples that was used to correct the TW of trawl samples: 
m Fr esh=-0-07561-f0.983697W Frozen (n=59, R 2 =0.996, 
P<0.001). When applied to data for trawl-caught rainbow 
smelt, these corrections increased post-frozen length and 
weight measurements by approximately 2^4% and decreased 
Fulton’s condition factor by approximately 4%. 
Length analysis Differences in TL among rainbow smelt 
from the 4 fyke net stations and from the trawl samples 
were evaluated for the study period by using the Kruskal- 
Wallis rank sum test and the Kruskal-Wallis multiple 
comparison test. Several trends were revealed from the 
analysis of pooled length data (Fig. 6). At all locations, males 
were shorter than females. Also, males and females from 
the southernmost station in the Jones River were shorter 
than those from all other locations, and males from the 
northernmost station in the Parker River were significantly 
longer than those from all other locations. Among samples 
from the trawl survey, the Fore River, and the Saugus River, 
there were no differences in TL except for females from the 
Fore River and Saugus River. This finding supports the 
premise that it was suitable to pool data from the trawl 
samples for the study area and compare length and age dis¬ 
tributions to data from samples collected in coastal rivers. 
This analysis and these length-frequency data raise no con¬ 
cerns over possible size selection that could have resulted 
from the use of trawl sampling gear (Fig. 2A). 
Analysis of differences in TL among aged rainbow smelt 
found significant differences beyond those revealed from 
the previously mentioned analysis of pooled length data. 
