Results Directory UMM :Data Elmu:jurnal:B:Biosystems:Vol56.Issue1.2000:

used since it alone was consistently derived from a non-spherical data set.

3. Results

The comparison of the degree of emergence, as AVGD, for the three different aged needles, young, middle age and old, are presented in Table 1. In no case is the difference between the sub- groups and the whole data set AVGD for any one age of needle primordia great, nor is differ- ence in AVGD among the differently-aged needle primordia great. Even so, all are statistically sig- nificantly different, even given a highly conserva- tive probability level for rejection of the null hypothesis of similarity, 0.0001. Table 2 presents the relationship between the degree of emergence, AVGD, and variation in size, SIZEVAR, and variation in integration, SHAPEVAR. This is a comparison similar to that presented in Maze 1999. Both estimates of varia- tion show a significant relationship with the de- gree of emergence AVGD, with variation in size having the larger absolute effect. In addition, the signs of the two estimates of variation are differ- ent, with the average degree of emergence declin- ing with variation in size but increasing with variation in integration. Table 3 shows the eigenvectors and percent variation accounted for by the first PCA axis for each entire data set describing young, middle age and older needle primordia, respectively. The pur- pose of this table is to offer insight into how the variable interrelationships change with age, giving more information on patterns of change as emer- gence occurs. The eigenvector elements for each needle variable measured, b width at base, m width at midpoint and l length, decline with age, with the greatest decline being in m and l. First axis eigenvalues also decline with age, an indication that the variables become less strongly polarized with age. Fig. 2 presents the notched box plots. All three variables become larger with age, an expected observation. There are also differences in time-re- lated variation as expressed in the size of the boxes and presence of outliers. The most variable Table 1 Average AVGD, standard deviations below, for young, middle age and old needle primordia a Old Young Middle age 1.1 2.8 0.4 0.1 0.6 0.5 a All are significantly different at PB0.0001. Table 2 Results of multiple regression analysis a Variable Coefficient Prob − 1.151 SIZEVAR 0.000 SHAPEVAR 0.913 0.000 a AVGD, average degree of emergence; SIZEVAR, varia- tion of needle primordia as measured by variance in PCA axis scores for young, middle age and old needle primordia; SHAPEVAR, variation in integration for each tree as mea- sured by variance in loadings on first eigenvector. First axis r 2 = 0.834. Table 3 Eigenvalues and eigenvectors for young, middle age and older needle primordia a Old Young Middle age Eigenvalues 1.5 1.7 2.4 0.819 0.796 b 0.778 m 0.697 0.935 0.766 l 0.724 0.910 0.655 a b, width at base; m, width of needle primordia at midpoint; l, length of needle primordia. measurements for width at the base of the needle primordia b are in the oldest largest amount of central variation defined by the notched box and youngest most outliers expressed needles. The least amount of variation for b is seen in the middle age needle primordia. For the width at the midpoint m and the length of the needles l, the most variable primordia were the youngest, with the middle age and older primordia appearing less variable. The patterns seen in Fig. 2 illustrate the level of within-structure needle variation where the greater variation in the youngest primordia grad- ually becomes more constrained through time. This constraint may be related to the loss of potential that accompanies development. At a more abstract level, this constraint may result as growing needles access information from a higher level phylogenetic clock. This ‘higher level infor- mation’ may be the reason why differences among species become more pronounced at later develop- ment stages. These changes in variation, while based on notched box plots, were confirmed by Bartlett’s test for homoscedasity Wilkinson, 1988 and co- efficients of variation. The results of those tests were not included.

4. Discussion