Interpreting the Legend from the RTA diagram:
Geographic Predictors: Follow the branches of the tree diagram
to the class number at the bottom of each node to match the legend for
this map. This output is a main advantage to the regression tree analysis
approach, as the map shows how different variables can drive the IV of
a species at different parts of its range.
The boolean expression is true if you follow the left
branch of the diagram. Also note that the length of the branch-arm is dependent on the change
of the parent and child nodes and denotes the importance of the split (longer the length,
more important the split)
Tree Diagram: This diagram is the output from the regression
tree analysis and is the basis for prediction of current and potential
future distributions of importance value. By following a branch with the
current (or predicted future) variables for a county, one can predict the
importance value for the species (shown at the terminal node of the branch,
along with the number of counties for that branch and the class for mapping).
The bracketed numbers at the bottom is the legend for the map of geographic
predictors. The n= numbers depict how many counties fall in that class. Variables that operate at large scales are depicted at the
top of the tree (e.g., broad climatic patterns), while more local effects
determine the local distributional variations.
R-Square: The splitting criterion for each node is the same as for linear
regression SST - (SSL + SSR), where SST
= (yi - ym)2 is the sum of squares
for the node and SSL and SSR are sums of squares for the right
and left branches, respectively. This is equivalent to chosing the split to maximize
the between-groups sums of squares in ANOVA. So the R2 is similar to that
defined in a linear regression and would denote the best acheived fit for the chosen
number of splits based on the above criteria.
facilitate mapping, we restricted the number of classes to nine. All values
below the chosen cutoff for the species is assigned to class 1 (species
is absent) whereas classes greater than 1 denote presence and increasing
importance of the species.