. The Canadian field-naturalist. Natural history. 196 The Canadian Field-Naturalist Vol. Ill. -1 0 1 Factor One Figure 1. 70% confidence ellipses for all nine species. Factor one, first PCA axis; Factor two, second PCA axis; o, Glyceria occidentalis; b, G. borealis; m, G. maxima; 1, G. leptostachya; g, G. grandis; p, G. pul- chella; c, G. canadensis; e, G. elata; s, G. striata. eliminate those variables whose vectors do not close- ly parallel major axes of variation; it is an approach designed to eliminate "noise" in the data. Three PCAs were performed, one for all species, one com-
. The Canadian field-naturalist. Natural history. 196 The Canadian Field-Naturalist Vol. Ill. -1 0 1 Factor One Figure 1. 70% confidence ellipses for all nine species. Factor one, first PCA axis; Factor two, second PCA axis; o, Glyceria occidentalis; b, G. borealis; m, G. maxima; 1, G. leptostachya; g, G. grandis; p, G. pul- chella; c, G. canadensis; e, G. elata; s, G. striata. eliminate those variables whose vectors do not close- ly parallel major axes of variation; it is an approach designed to eliminate "noise" in the data. Three PCAs were performed, one for all species, one com- paring G. canadensis and G. pulchella and one com- paring G. elata and G. striata. The results of the PCA, the PCA axis scores, were used to produce scatter plots, summarized as confidence ellipses, and subjected to analysis of variance (ANOVA) to offer a quantitative summary of the scatter plots. For this we relied on Tukey's multiple range test. In all the comparisons using PCA axis scores we used only the first two axes as these are the ones that account for most the varia- tion in the data. These and subsequent analytical comparisons were done using SYSTAT Network Version Wilkinson (1991). Results Analysis of all species: The first PCA, done with all the variables for all species, resulted in eliminating six variables, BLADEL, BLADEW, BRANCH, NODE, SPIKW and SAPEX. In the PCA on the remaining variables (Table 3), the first axis of the PCA explained of the data variation. The characters load- ing heaviest on this axis pertained to the spikelet and floret characteristics; the four most important were: PALEA, SPIKL, LEMMA and GL2. However all variables have fairly high loading and the same sign indicating the first axis most likely reflects size differences. The second axis explained of the data variation and the most impor- tant characteristics were LIGULE, GLl, SHEATH and ANTHER. The presence of both positive and negative signs in the loading on the second
Size: 1588px × 1573px
Photo credit: © Book Worm / Alamy / Afripics
License: Licensed
Model Released: No
Keywords: ., bookcentury1900, bookdecade1910, booksubjectnaturalhistory, booky
