Results and discussions
Regression tree technique has failed to read the data structure due to
collinearity effect. The so-called “statistical hazards” is identified
from the extremely high correlation coefficients (R2):
ec vs B, Na, Cl; Na vs K, Li; NH4 vs K, Li. Only three
branches are extracted from the procedure, involving air temperature
(atemp), elevation (elv), and B (Boron concentration), which is very
difficult to explain (Figure 2). However B is presumably has the
strongest control to the data structure as it closely correlates with EC
along with Na and Cl. Li and K both have strong correlation with Na and
NH4.
Table 3 A correlation diagram showing several variables that has
extremely high (0.98-0.99) R2