| About REGRESS.pro [message #47804] |
Tue, 07 March 2006 08:37  |
Daelomin
Messages: 7
Registered: March 2006
|
Junior Member |
|
|
Hi,
I discovered the newsgroup existence from a post on David Fanning's
website. Thanks sir!
I am working on a somewhat simple multi-linear regression problem from
brightness temperatures of a satellite to cloud liquid water.
As I have started to investigate the REGRESS procedure in IDL, I just
realized that I don't quite know its code. Is it totally kosher by
Numerical Methods standards?
Does anyone have another version that I could use to at least compare
the validity of outputs?
Thanks mucho,
Rémi
|
|
|
|
| Re: About REGRESS.pro [message #47875 is a reply to message #47804] |
Wed, 08 March 2006 06:31  |
K. Bowman
Messages: 330
Registered: May 2000
|
Senior Member |
|
|
In article <1141808047.118961.137550@z34g2000cwc.googlegroups.com>,
"Daelomin" <joey.pourriciel@gmail.com> wrote:
> Very nice!
>
> IDL> stochastic_test(n=1000)
> 1.00000 0.500000
> 1.00255 0.526448
> IDL> stochastic_test(n=10000)
> 1.00000 0.500000
> 0.998914 0.493785
> IDL> stochastic_test(n=100000)
> 1.00000 0.500000
> 0.999659 0.500338
>
> Convergence rate is a bit slow it'd seem no? What do think Ken?
>
> All in all it's ok, just means you need really large datasets I guess.
Most introductory statistics books discuss the sampling distributions of the
linear regression estimates for a, b, and r.
Also, I would include the /DOUBLE keyword in REGRESS to help minimize
accumulated roundoff error.
Ken Bowman
|
|
|
|
| Re: About REGRESS.pro [message #47880 is a reply to message #47804] |
Wed, 08 March 2006 01:46 |
Paolo Grigis
Messages: 171
Registered: December 2003
|
Senior Member |
|
|
Daelomin wrote:
> Hi,
>
> I discovered the newsgroup existence from a post on David Fanning's
> website. Thanks sir!
>
> I am working on a somewhat simple multi-linear regression problem from
> brightness temperatures of a satellite to cloud liquid water.
>
> As I have started to investigate the REGRESS procedure in IDL, I just
> realized that I don't quite know its code.
The source code is available in the lib subdirectory of your idl
distribution.
Ciao,
Paolo
Is it totally kosher by
> Numerical Methods standards?
>
> Does anyone have another version that I could use to at least compare
> the validity of outputs?
>
> Thanks mucho,
>
> R�mi
>
|
|
|
|
| Re: About REGRESS.pro [message #47881 is a reply to message #47804] |
Wed, 08 March 2006 00:54 |
Daelomin
Messages: 7
Registered: March 2006
|
Junior Member |
|
|
Very nice!
IDL> stochastic_test(n=1000)
1.00000 0.500000
1.00255 0.526448
IDL> stochastic_test(n=10000)
1.00000 0.500000
0.998914 0.493785
IDL> stochastic_test(n=100000)
1.00000 0.500000
0.999659 0.500338
Convergence rate is a bit slow it'd seem no? What do think Ken?
All in all it's ok, just means you need really large datasets I guess.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
