Uncertainty, Emergence, and Statistics in Dendrochronology - Dendroclimatology - page 91

A. 1 EIGENVECTOR -- R 2 = 0.195 (32 sig.)
0.1
0
-0.1
B. 2 EIGENVECTORS -- R 2 = 0.250 (26 sig.)
0.1
0
-0.1
C. 3 EIGENVECTORS -- R 2 = 0.303 (27 sig.)
0.1
0
-0.1
D. 4 EIGENVECTORS -- R 2 = 0.342 (24 sig.)
0.1
0
-0.1
MONTHS
Fig. 4.4 The stepwise development of the hemlock response function shown in Fig. 4.3b . The
progressive changes in fractional variance explained ( R 2 ) and the number of significant coefficients
are indicated
is highly sensitive to (1) the number of eigenvectors retained as candidate predictors
in regression analysis and (2) the criterion for entering eigenvectors into the regres-
sion model. Fritts ( 1976 ) , Guiot et al. ( 1982 ) , and Fekedulegn et al. ( 2002 ) argue
for retaining a large number of candidate eigenvectors, ones that may explain up to
90-95% of the total variance in the climate data correlation matrix. An objective