de Silva and Condrey: Patterns in patchy data discerned from Brevoortia patronus bycatch 
207 
® 
E 
b 
1 - 
ab 
but 
q 
bs 
tf 
shk 
AM5&16 
ac 
acu 
AM7&18 
spm 
bts 
cr 
rd gc 
-he- 
-i -i 
AM1&12 
bsk, 
sh 
sps 
stm 
-i 
Dimension 2 
Figure 8 
Correspondence analysis plot for species-area associations (summer) for 1995 fishing season. 
find factors that could explain the bycatch, and we 
used loglinear models to find a suitable model that 
was associated with occurrences of bycatch greater 
than the median level for the fishery. 
By using both loglinear models and logit models 
with categorical explanatory variables, we have used 
the loglinear model for model selection and the logit 
model for a detailed examination of the model of our 
choice in a manner analogous to AN OVA. 
With the stepwise selection procedure, [SAB SAD] 
was found to be the most suitable loglinear model. We 
had hypothesized that the presence of dolphins in the 
fishing area could indicate the presence of high bycatch. 
Had this been true, the presence of dolphins in the vi- 
cinity could have been used by fishermen to avoid set- 
ting nets in certain areas. However, such a relation 
between dolphins and bycatch was found not to exist. 
At first glance the issue of bycatch in Gulf of Mexico 
menhaden fishery may seem to be negligible, given 
the low bycatch percentage. However the fishery had 
the second highest annual U.S. commercial landings 
of 472,000 t in 1995 (U.S. Dep. Commerce, 1996). 
Further, given the strong positively skewed distri- 
bution of releasable bycatch, a small percentage of 
the total fishing effort would account for much of the 
take. By structuring our analysis around bycatch 
rates that were greater than the median, we have 
attempted to identify the potential “hot spots” in the 
fishery in terms of areas and seasons. One of the so- 
lutions for reducing bycatch in the fishery is to iden- 
tify such areas and thereby offer the industry a tool 
for managing their take of bycatch by minimizing 
fishing effort in these “hot spots.” 
Our philosophy has been that consideration of all 
bycatch as a single entity is not the best approach. 
Although we felt the need to address the total bycatch 
in our first analysis, we also wanted to take indi- 
vidual species into consideration. We approached this 
multispecies aspect of our study using correspon- 
dence analysis, which we used to identify the com- 
monly associated species in different zones and ar- 
eas, as well as the different fates. Our results sug- 
gest that this approach can have general appeal not 
only in identifying areas and species of concern but 
also in suggesting approaches to solutions. For ex- 
ample, our “hot spot” fishing zone and season were 
