WILKINS ET AL: FLORIDA PANTHER MORPHOLOGY


 CDA is a powerful procedure that maximizes intergroup differences to
portray the relationships of the groups more clearly. CDA utilizes OTUs that have
been divided into groups on the basis of an a priori classification. It may be used
to assign group membership to new specimens, or to describe group differences
and relationships. MANOVA emphasizes the testing of similarity/difference
among centroids of the a priori groups and, in that respect is closely related to
discriminant analysis. MANOVA tests the hypothesis that all a priori groups have
the same multi-dimensional mean centroidd) for the variables measured.
 In order to examine the morphometric relationships within the Florida
population, where a priori classification was required, the Florida and southeast
specimens were assigned to the following classes: Florida historic (=HIST, pre-
1950 Florida specimens); Florida recent (=RECENT, non-Everglades Florida cats
since 1950); historic Louisiana specimens (=ARUND); Everglades cats
gladesES); animals from Florida of questionable identity, or specimens with no
data (=TEST), Piper captive cats from Everglades Wonder Gardens (=PIPER); and
the southeastern cats from Arkansas and Louisiana (=ARK/LA) that were collected
between 1965 and 1975 long after cougars were thought to survive in Louisiana
(Goertz and Abegg 1966, Sealander and Gipson 1973). The TEST animals are two
females shot in Palm Beach County outside the known range of the Florida
population, a specimen in the Everglades Regional Collection Center with no data,
the skull of a male cougar found in Volusia County, a male from South Carolina
with mixed data, and a specimen from a private collection recently donated to the
Florida Museum of Natural History with no data. These labels are used throughout
the following discussion. At one time, the canebrake puma from Louisiana was
considered a separate subspecies P. c. arundivaga (Hollister 1911), but Nelson and
Goldman (1929) synonomized it with P. c. coryi since they were unable to find any
distinctive characters to separate it.
 Only adults were included in the study. Cougars are sexually dimorphic, with
 males being larger than females (Goldman 1946, Kurtdn 1973, Anderson 1983,
 Maehr and Moore 1992, Gay and Best 1995). This sexual variation dictates
 separate analyses by sex, at least for variables associated with skull measures,
 thereby reducing the effective sample size for each subspecies. Skins of juveniles
 and those that were notably faded as a result of continuous exposure to light were
 eliminated from the color analysis. All data sets were tested for normality prior to
 analyses using the Shapiro-Wilk statistic (W). With one exception (cranial
 profile), multivariate analyses were done with the Statistical Analysis System (SAS
 Institute Inc. 1985). Group sample sizes varied with each statistical procedure,
 depending upon the availability and condition of specimens. Many skulls were
 damaged or did not have skins, so it was impossible to combine variables, since
 complete specimens (skins, or undamaged skulls) were often unavailable. Also,
 some characters were qualitative and others quantitative. Therefore, each
 character analysis was conducted independently. Because sample sizes and