1. Forum
  2. Usenet
  3. COMP.GRAPHICS.APPS.GNUPLO

  • strange discrepancy between stats command result and fitting result

    From [email protected]@21:1/5 to Karl Ratzsch on Sun Apr 14 23:46:11 2019
    On Sunday, April 14, 2019 at 6:56:16 PM UTC+2, Karl Ratzsch wrote:
    Am 14.04.2019 um 17:19 schrieb [email protected]:
    I am puzzled by the following discrepancy between a stats command result and the result which I have from normal Gnuplot fitting:

    f(x)=a*log10(x)+b
    a=-1.06
    b=-4.78
    fit [3e-3:1] f(x) datafile using ($1):($2) via a,b
    stats [3e-3:1] datafile using (log10($1)):($2)


    The Gnuplot fitting gives me y=ax+b, with the correct a=-1.16816 and b=-4.78604
    But the stats-command gives: y = -3.104 x - 4.194


    Not problem with fit or stats, your range setting for stats was
    wrong: you have to logarithmise it as well

    stats [log10(3e-3):log10(1)] $dat using (log10($1)):($2)

    and the linear model that comes out is y = -1.168 x - 4.786

    ;-)

    Thanks a lot!
    kind regards,
    Hugo

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)