[Gnuplot-bugs] bugs using fit

[Wanderson W. <wan...@if...>]

Dear,


I tried using fit function in gnuplot version 4.0 patchlevel 0 (linux debia=
n),=20
but something strange happen.

=46rist I define a function
gnuplot> g(x) =3D a*exp(-b/x) +c*exp(-2*b/x)

and make the fit
gnuplot> fit [1:16] g(x) 'omegQ_25bt.dat'  via a,b,c
=2E . . . .=20
degrees of freedom (ndf) : 13
rms of residuals      (stdfit) =3D sqrt(WSSR/ndf)      : 2.17102e-05
variance of residuals (reduced chisquare) =3D WSSR/ndf : 4.71334e-10

=46inal set of parameters            Asymptotic Standard Error
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D      =
      =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D

a               =3D 0.000550667      +/- 0.001319     (239.5%)
b               =3D 27.5771          +/- 4.699        (17.04%)
c               =3D 0.0268601        +/- 0.009831     (36.6%)


correlation matrix of the fit parameters:

               a      b      c     =20
a               1.000=20
b               0.982  1.000=20
c               0.936  0.985  1.000=20

The file omegQ_25bt.dat have 3 data columns:  x, f(x), error in f(x) , then=
 I=20
want use this error, I put "using 1:2:3" in commannd line

gnuplot> fit [1:16] g(x) 'omegQ_25bt.dat' using 1:2:3 via a,b,c
degrees of freedom (ndf) : 13
rms of residuals      (stdfit) =3D sqrt(WSSR/ndf)      : inf
variance of residuals (reduced chisquare) =3D WSSR/ndf : inf

=46inal set of parameters            Asymptotic Standard Error
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D      =
      =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D

a               =3D 0.000550667      +/- nan          (nan%)
b               =3D 27.5771          +/- nan          (nan%)
c               =3D 0.0268601        +/- nan          (nan%)

correlation matrix of the fit parameters:

               a      b      c     =20
a                 nan=20
b                 nan    nan=20
c                 nan    nan    nan=20

OK!!! is no possible make fit that way, than I put off  "using 1:2:3",=20
however, the results change:=20

gnuplot> fit [1:16] g(x) 'omegQ_25bt.dat' via a,b,c

degrees of freedom (ndf) : 13
rms of residuals      (stdfit) =3D sqrt(WSSR/ndf)      : nan
variance of residuals (reduced chisquare) =3D WSSR/ndf : nan

=46inal set of parameters            Asymptotic Standard Error
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D      =
      =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D

a               =3D nan              +/- nan          (nan%)
b               =3D nan              +/- nan          (nan%)
c               =3D 0.0268601        +/- nan          (nan%)


correlation matrix of the fit parameters:

               a      b      c     =20
a                 nan=20
b                 nan    nan=20
c                 nan    nan    nan=20


This is a bug, I think,


That's all

Thanks


=2D-=20
=2D-------------------------------------------------------------------

Wanderson Gon=E7alves Wanzeller

Instituto de F=EDsica Te=F3rica - IFT / UNESP

Rua Pamplona, 145=20
CEP: 01405-900
S=E3o Paulo - SP