Closed qichun-dai closed 2 years ago
Thank you Qichun!
Yes I agree with you. I was a bit lazy and only wrote "Similar arguments apply to the general m" for brevity :(
Haha, I see :)
On Tue, Apr 20, 2021 at 12:14 AM Yuhang @.***> wrote:
Thank you Qichun!
Yes I agree with you. I was a bit lazy and only wrote "Similar arguments apply to the general m" for brevity :(
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/YuhangZhou88/ESL_Solution/issues/3#issuecomment-822594474, or unsubscribe https://github.com/notifications/unsubscribe-auth/ALZ53FGUBWXWYMOSCRPUMSDTJRJHHANCNFSM43F3KTAQ .
-- Best, Qichun
Hello, recently I am also working on this exercise (https://github.com/szcf-weiya/ESL-CN/issues/236). It seems that the partition region R_m
depends on y_i
, so the covariance you calculated is just a conditional one. (I am struggling in the closed-form without fixing R_m
, and by googling related materials, I found your great repo here!
Your code indeed treats R_m
as random, since you fit a new regression tree for each column of y
, but there is a typo that might cause the estimation is close to the number of terminal nodes,
the summation is over each observation, i.e., along the row.
@szcf-weiya Thank you! you are correct, that should be 0.
For (a), I feel what the authors ask is the conditional one, from the context. Dof is a complicated concept for models like trees.
@YuhangZhou88 To my best knowledge, I had not heard of "conditioanl dof" in any related papers. If you find one, I am appreciated if you can share its link.
@szcf-weiya I mean the variance, as you pointed out, is "a conditional one". My second comment on dof is a general comment, so for the "closed-form without fixing R_m" that you are looking for, I can imagine it's difficult.
Hi, Yuhang: Thanks for developing this site for ESL!
I learned how to solve 9.5 (a) from your solution. You can probably generalize the proof as follows.
Best, Qichun