Re[2]: TS prescision - question - non-linear regression

[Jim Johnson <jejohn@xxxxxxxxx>]

Hello list,

I could be wrong--happens quite a lot actually--but I think Dennis
Meyer's work uses curvilinear methods.  His systems are explained at
www.meyersanalytics.com and he has written often in TASC in recent
years.



Best regards,
 Jim Johnson   

Thursday, August 02, 2001, 8:39:41 PM, you wrote:


TT> Prosper,

TT> Linear regression is a least squares fit of a line (ax+b) to a window of
TT> data.

TT> Non linear regression could mean:

TT> 1) A linear fit of a set of functions (which themselves could be non-linear)
TT> -or-
TT> 2) A least squares error fit of higher order polynomials (quadratic,
TT> cubic...)
TT>     (Actually this is a special case of #1).

TT> The Savgol routine in Numerical Recipes does #2 if you are into C/C++.
TT> There are also routines in NRC to do #1.

TT> And yes, filters derived from the last point of a moving non-linear fit are
TT> quite useful in systems.

TT> No, I have not seen code to do these in TradeStation.

TT> Best,
TT> Tim Tillson

TT> ----- Original Message -----
TT> From: "Prosper" <brente@xxxxxxxxxxxx>
TT> To: "Omega List" <omega-list@xxxxxxxxxx>
TT> Sent: Thursday, August 02, 2001 7:08 AM
TT> Subject: TS prescision - question - non-linear regression


>> I have always wondered about non-linear regression ever since I heard it
>> mentioned some where on the Internet.
>>
>> First. does anyone have a easy language function for non-linear regression
>> that they wouldn't mind sharing?
>>
>> Second, if linear Regression is least squares, is non-linear regression
TT> most
>> squares?
>>
>> Third, is non-linear regression useful to any of you traders out there?
>>
>> Prosper
>>
>>



-- 
                        mailto:jejohn@xxxxxxxxx