193 
Abstract .—We use two categorical 
techniques to explore patchy data on re- 
leasable bycatch in the U.S. Gulf of 
Mexico menhaden fishery. With no pre- 
vious statistical studies on releasable 
bycatch, this paper serves both as an 
analysis of patchy data with categori- 
cal techniques and as a quantitative 
description of a biologically important 
portion of the menhaden bycatch. 
By means of exploratory analyses 
with loglinear and logit models, we de- 
termined spatial and temporal patterns 
in bycatch of the menhaden fishery. 
Contrasts revealed that at fishing 
grounds east of the Mississippi River, 
the probability of observing fishing sets 
with high bycatch rates in spring and 
summer were greater than in the fall. 
Furthermore, spring bycatch rates were 
higher in fishing areas east of 89°W 
than in fishing areas west of 93°W. 
Correspondence analysis indicated 
that the fate of the releasable bycatch 
could be classified into three major 
groupings. The first group, species as- 
sociated with being gilled, was com- 
posed primarily of Micropogonias 
undulatus, Trichiurus lepturus, Chloro- 
scombrus chrysurus, and Cynoscion 
arenarius. The second group consisted 
of species associated with being dead 
or disoriented; it included the requiem 
sharks Caranx hippos and Sciaenops 
ocellatus. The third group included 
those fish that were associated with be- 
ing put into the hold, those kept by the 
crew, or those whose fate was unknown. 
These included Arius felis , Bagre mar- 
inus, and Scomberomorus maculatus. 
Seasonal and spatial associations of 
bycatch species were also examined 
with correspondence analysis. From 
April through August, two distinct 
bycatch species assemblages were ob- 
served that separated the fishery at a 
longitude of 91°W. From September 
through October, a shift in the species 
assemblage indicated that the western 
region of the fishery (west of 93° W) 
appeared to have a assemblage distinct 
from the rest of the fishery. 
Manuscript accepted 18 August 1997. 
Fishery Bulletin 96: 193-209 ( 1998). 
Discerning patterns in patchy data: 
a categorical approach using gulf 
menhaden, Brevoortia patronus, 
bycatch 
Janaka A. de Silva 
Department of Oceanography and Coastal Sciences 
Louisiana State University 
Baton Rouge, Louisiana 70803 
E-mail address:Jandesilva@worldnet.att. net 
Richard E. Comdrey 
Department of Oceanography and Coastal Sciences 
Coastal Fisheries Institute, Louisiana State University 
Baton Rouge, Louisiana 70803-7503 
Statistical analyses of fisheries data 
often require consideration of the 
patchy distribution of aquatic or- 
ganisms (e.g. Andrew and Peppered, 
1992). This is especially true with 
standard regression and ANOVA 
techniques because they rely on the 
assumptions of normality of residu- 
als and homogeneous variances. 
Often, transformations of the re- 
sponse variable can be used to sat- 
isfy these assumptions (Underwood, 
1981). However, in many instances 
suitable transformations cannot be 
found. In such cases, the use of stan- 
dard regression and ANOVA tech- 
niques will result in gross violations 
of the model assumptions and affect 
the results of the analysis (Under- 
wood, 1981). Where suitable trans- 
formations cannot be found, non- 
parametric techniques such as the 
Kruskal-Wallis one-way ANOVA or 
ANOVA on the ranked data have 
been used. As Underwood (1981) 
has noted, one of the implied as- 
sumptions of these methods, that of 
equal variances among treatments, 
is usually not satisfied when deal- 
ing with patchy data. A further 
problem with such techniques is 
that one cannot explore significant 
differences with the use of contrasts 
or multiple comparison tests, except 
with the Kruskal-Wallis one-way 
ANOVA. Therefore, when using 
these techniques, one is limited to 
the use of empirical methods to ex- 
amine significant differences. 
Studies on fisheries bycatch have 
used a variety of statistical meth- 
ods. Andrew et al. (1995) used 
ANOVA with suitable transforma- 
tions to evaluate the bycatch in an 
Australian stow-net fishery for 
school prawns, and Austin et al. 
( 1994) used the Kruskal-Wallis one- 
way ANOVA to examine geographi- 
cal differences in the bycatch of the 
Atlantic menhaden fishery. Hudson 
(1990), examining the shrimp and 
fish bycatch assemblages of the 
Canadian Eastern Arctic, used an 
intermediate linkage clustering al- 
gorithm to examine abundance pat- 
terns and species associations 
among different regions. More 
unique solutions have been pre- 
