Home > Archive > APL > February 2006 > Levenberg-Marquardt / 4 parameter Logistic
You are viewing an archived Text-only version of the thread.
To view this thread in it's original format and/or if you want to reply to
this thread please [click here]
| Author |
Levenberg-Marquardt / 4 parameter Logistic
|
|
| LarryM 2006-02-24, 6:56 pm |
| This topic last came up in in 2001, but nothing was ever resolved.
I need to fit a 4 parameter logistic model to some data and would like
use a Levenberg-Marquardt algorithm written in APL+Win. The program
should accept x and y values, calc the 4 Ps, some estimated values, and
some measure of goodnes of fit. There is no need for GUI, graphics,
fancy formatting, or additional features.
Larry Mysz
Customized Applications
| |
|
|
| valter.sundh@geriatrik.gu.se 2006-02-26, 6:55 pm |
| If you want an APL program that calculates the parameters of a binary
logistic regression model, you can find that in Vector vol. 14 no. 2
October 1997. It uses weighted least squares to find the solution, but
does it really matter what method is used, as long as the correct
solution is found?
Valter Sundh
Goteborg university
LarryM skrev:
> This topic last came up in in 2001, but nothing was ever resolved.
>
> I need to fit a 4 parameter logistic model to some data and would like
> use a Levenberg-Marquardt algorithm written in APL+Win. The program
> should accept x and y values, calc the 4 Ps, some estimated values, and
> some measure of goodnes of fit. There is no need for GUI, graphics,
> fancy formatting, or additional features.
>
> Larry Mysz
> Customized Applications
| |
| Björn Helgason 2006-02-26, 6:55 pm |
| Looking for Levenberg-Marquardt apl on google gave 188 hits
| |
| LarryM 2006-02-27, 9:55 pm |
| Thank you to all who responded both here and off list.
I do want to look at the 2PL program in Vector, but I cannot use it in
this application. This one really does need a 4PL model.
One piece of my application determines calibration curves used to
estimate results of tests run in a clinical lab. The maker of one test
kit has specified a 4PL model. So, in this one case, the correct
solution is a 4PL solution by definition.
I want the Levenberg-Marquardt program rather than something
specifically 4PL because L-M can be used for other non-linear curve fit
problems that I run across every now and again.
Re google searches for Levenberg-Marquardt in APL ... in about 190 of
the 200 hits, APL is part of Applied Physics Lab or Journal of Apl
Math.
One off list response suggested I check the ACM publications of APL
Conference the Proceedings. I found one L-M model in APL94, but those
programs are copyright by ACM, and I need distribution rights.
I had hoped to find a tested, ready to run, APL version for sale, but
it doesn't look like one is readily available. So I have started
translating the version from Numerical Recipes in FORTRAN into APL. I
am still be interested in a L-M program ... at least until that
translation is done and tested.
Larry Mysz
Customized Applications
|
|
|
|
|