single index model

13 views
Skip to first unread message

josef...@gmail.com

unread,
Jan 7, 2013, 6:02:20 PM1/7/13
to pystatsmodels
(just showing a plot, blogging has to wait)

y = g(x * b) + e

proof by plotting, maybe "it works"

Josef
ex_singleindex.png

Ralf Gommers

unread,
Jan 9, 2013, 3:58:25 PM1/9/13
to pystat...@googlegroups.com
Little wiggles at the boundaries of the support look odd, other than that it looks pretty good.

Ralf
 

josef...@gmail.com

unread,
Jan 9, 2013, 4:16:55 PM1/9/13
to pystat...@googlegroups.com
It might be mainly because there are fewer observations at the ends.

(adaptive bandwidth again ?)

Josef

>
> Ralf
>
>

Ralf Gommers

unread,
Jan 10, 2013, 2:27:43 AM1/10/13
to pystat...@googlegroups.com
Possibly. IIRC this is also a known issue occurring with local constant regression; local linear should be better in that respect. No time to look up now what SingleIndex is using, so this comment may be irrelevant.

Ralf
 
Reply all
Reply to author
Forward
0 new messages