On one hand, in the section 4.1, page 13 of the pedf, it is said:
Using the first result, we can uniquely define $T\alpha$ as the smallest tree $T$ for which $R\alpha (T) $ is minimized.
My question is if shouldn't be the biggest tree instead of the smallest, since, as I understand, it is around a sub tree of the full model. Am I wrong?
Secondly, just after that the intervals are printed as:
Looking at the table, we see that the best tree has 10 terminal nodes (9 splits), based on cross-validation.
And then it is claimed that
This sub tree is extracted with a call to prune and saved in fit9.
However the prunefit9 extracted has 10 splits (and 11 terminal nodes, as in Figure 4), as it uses cp = 0.2 > 0.022222, and so, with the notation of the intervals (with my correction[???]), this cp = 0.2 belongs to $I_5 = [0.0166667, 0.0222222)$.
On one hand, in the section 4.1, page 13 of the pedf, it is said:
My question is if shouldn't be the
biggest tree
instead of thesmallest
, since, as I understand, it is around asub tree of the full model
. Am I wrong?Secondly, just after that the intervals are printed as:
However, brackets seem to be reversed. Shouldn't they be as follows?
Finally, in section 4.3 it is found that
And then it is claimed that
However the
prune
fit9
extracted has 10 splits (and 11 terminal nodes, as in Figure 4), as it usescp = 0.2 > 0.022222
, and so, with the notation of the intervals (with my correction[???]), thiscp = 0.2
belongs to $I_5 = [0.0166667, 0.0222222)$.Thanks!