apologies if this is a very simple question, but I have searched everywhere and can't find a solution. I have a data file with three columns - x:y:xerrors. I want to fit a function f(x) to the data, by entering
f(x)=a*x+b
fit f(x) 'data' using 1:2:3 via a,b
But is this fit assuming that the entries in column 3 are y-errors, when actually they are x-errors?
Change: fit f(x) 'data' using 1:2:3 via a,b
To: fit f(x) 'data' using 1:2:3 with xerrorbars via a,b
This treats the 3rd column as the uncertainty in x and computes an appropriate weight for the WSSR.
Yes the fit assumes that the 3rd column is y-error if you specify
using 1:2:3 else (if you do NOT specify) it assumes all weight equal to 1.
Meanwhile we wait the patch will be included in all the gnuplot distribution,
I can propose a Workaround: # y=ax+b --> x=1./a*(y-b)
f(x)=a*x+b
f2(x)=1./a * (y-b)
fit f2(x) 'data' using 2:1:3 via a,b
plot 'data' with xerrorbar, f(x)
of course this changes the meaning of the uncertainty associated with a and b. See here.
