Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.graphics.apps.gnuplot > #316

Re: Multible Fits using the same fit function lead to ?false? results.

Show all headers | View raw

Hi,
On 18.05.2011 00:04, Oliver Götze wrote:
> i just noticed that using a gnuplot script like this:
> fit a0 + a1*(1/x)**2 + a2*(1/x)**4# table.dat  |awk "\$1 =="10" {print 1,\$2} \
> \$1 =="10" {print 2, \$3}\
> \$1 =="10" {print 3, \$4}\
> \$1 =="10" {print 4, \$5}"'via a0,a1,a2
> print "value a0 ",a0
>
> fit a0 + a1*(1/x)**2 + a2*(1/x)**4\
> '<grep -v \# table.dat  |awk "\$1 =="20" {print 1,\$2} \
> \$1 =="20" {print 2, \$3}\
> \$1 =="20" {print 3, \$4}\
> \$1 =="20" {print 4, \$5}"'via a0,a1,a2
> print "value a0 ",a0
>
> on a table.dat file like this (just an example):
> #x     y1     y2     y3
> 10     12     43     42
> 20    234    324   33
> 30     3         3       3
>
> leads to false results, i have to reset (or set to zero) die
> coefficients a0, a1, a2 after/before every fit by placing

Just to make it clear: gnuplot needs initial values for the fitting 
parameters. If they are not defined, they will be created with an 
initial value of 1.0 (see manual). Of course, you can preset the 
parameters with more adequate initial values for you problem. So, how 
can gnuplot know by itself, that the values of the fitting parameters 
are wrong and that they should be reset?
Of course you must reset by yourself with an explicit
a0=1.0; a1=1.0; a2=1.0;
between two independent fits.

Christoph

Back to comp.graphics.apps.gnuplot | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

Multible Fits using the same fit function lead to ?false? results. Oliver Götze <goetze.oliver@googlemail.com> - 2011-05-17 15:04 -0700
  Re: Multible Fits using the same fit function lead to ?false? results. Christoph Bersch <usenet@bersch.net> - 2011-05-18 09:54 +0200
  Re: Multible Fits using the same fit function lead to ?false? results. Hans-Bernhard Bröker <HBBroeker@t-online.de> - 2011-05-19 01:56 +0200

csiph-web