Laserfiche WebLink
<br />. <br /> <br />.- <br /> <br />replaced by 95 percent confidence ellipses, to allow the unambi- <br /> <br />guous display of larger data sets (Winans, 1985). <br /> <br />being developed by the author for these purposes.' <br /> <br />SofbJare is <br /> <br />PCJ\.-axis <br /> <br />values for each fish were displayed on bivariate scatterplots for <br /> <br />the direct visual examination of patterns within the morphometric <br /> <br />variation. <br /> <br />RESULTS <br /> <br />Ninety-nine percent of variation was accourH:~d by 19 PCA axes. <br /> <br />The first thr(~e axes explained 87% of the total variance: 76% for <br /> <br />axis 1 (PC1), 7% for axis 2 (PC2), and 4% for axis 3 (PC3). <br /> <br />Other axes individually accounted for 2% or less variation and <br /> <br />were excluded from further consideration. <br /> <br />The character loadings or scores indic~ted the dAgree to which <br /> <br />measurements contributed to the variance in each PCA axis, with <br /> <br />loadings considered significant if they exceeded an absolute <br /> <br />value of 0.30 (out of a possible 1.00; Dunn and Everitt, 1982) <br /> <br />Lo~dings for the first 3 principal component axes are presented <br /> <br />in Table 1. <br /> <br />All PC1 scores were significant, with values <br /> <br />exceeding +0.60. <br /> <br />This axis was considered a general indicator of <br /> <br />size. <br /> <br />Standard length, the traditional size indicator, hac a <br /> <br />score of +0.98 and this character tended to increase from left to <br /> <br />right along PC1 (Fig. 2). <br /> <br />A strong association between size as a <br /> <br />general f~ctor and PC1 is a usual outcome of PCA's involving <br /> <br />6 <br />