. Bulletin of the Museum of Comparative Zoology at Harvard College. Zoology. — .32 -- .38 .44. Figure 8. Third order trend surface analysis (with interaction terms suppressed) for projection of Grand Ba- hamian samples on the first varimax axis of a Q-mode analysis; this is a "size" axis based on all variables. Note simple pattern of increasing size from north to south, with more rapid transition near the northern coast, where dwarfed samples pass rapidly to interior samples of modest size. Actual localities indicated by crosses. This smooth variability has, in the past, been parcell
. Bulletin of the Museum of Comparative Zoology at Harvard College. Zoology. — .32 -- .38 .44. Figure 8. Third order trend surface analysis (with interaction terms suppressed) for projection of Grand Ba- hamian samples on the first varimax axis of a Q-mode analysis; this is a "size" axis based on all variables. Note simple pattern of increasing size from north to south, with more rapid transition near the northern coast, where dwarfed samples pass rapidly to interior samples of modest size. Actual localities indicated by crosses. This smooth variability has, in the past, been parcelled among three separate species defined only by differ- ences in size. "goodness of ; The "art" of trend surface analysis involves the selection of a fit that explains enough information, yet remains sufficiently simple to represent a truly re- gional pattern. Points can be fit exactly with polynomial surfaces of sufficiently high order. We used the program of Lee (1969). We decided not to use the mean of in- dividual characters as dependent variables, but a value expressing major determinants of covariance among samples. Consequently, we performed a Q-mode factor analysis of all C. bendaJU samples from Grand Bahama and used loadings on the first varimax axis (for a three-axis solution) as the dependent variable. This single axis encompasses per cent of the variance among 20 characters for the 12 samples. Factor scores of vari- ables upon it (Table 3) show that it repre- sents a fairly "pure" size axis, with high and similar loadings for measures of final size and whorl size. (We do not detect the com- mon negative interaction here, because we do not consider alternate pathways to a similar final size. We have, instead, the opposite situation—a wide range of mean shell size from very small on the north coast to quite large elsewhere. The dwarfed shells have both few whorls and small whorls.) The first order fit alone has a multiple cor
Size: 3491px × 716px
Photo credit: © Book Worm / Alamy / Afripics
License: Licensed
Model Released: No
Keywords: ., bookauthorharvarduniversity, bookcentury1900, booksubjectzoology
