4

Intro

In gnuplot there's a solution to create histogram from file named hist.dat what likes

1
2
2
2
3

by using commands 

binwidth=1
set boxwidth binwidth
bin(x,width)=width*floor(x/width) + binwidth/2.0
plot [0:5][0:*] "hist.dat" u (bin($1,binwidth)):(1.0) smooth freq with boxes

that generates a histogram like this one from other SO page.

Question

How can I fit my function to this histogram? I defined a Gaussian function and initialized its values by

f(x) = a*exp(-((x-m)/s)**2)
a=3; m=2.5; s=1

and in the output the function follow the histogram well.

Unfortunatelly I cannot fit to this histogram using command

fit f(x) "hist.dat" u (bin($1,binwidth)):(1.0) smooth freq via a,m,s  
                                                      ^
         Need via and either parameter list or file

So how can I fit my function without creating a new file containing the binned values?

Community
  • 1
  • 1
DanielTuzes
  • 2,494
  • 24
  • 40
  • 1
    Although the question is quite old: Why don't you want a new file? You can delete that file at the end of the script: `system('del hist.dat')` or `system('rm hist.dat')`. – Christoph Sep 30 '13 at 17:06
  • Inefficient and OS-specific. If I want to generalize my solution to larger data, writing and reading a file can become the bottleneck and makes the source code longer. If I want to show the script to colleagues, using an OS-specific solution is never a good idea. Nowadays I think python suits more in many cases. – DanielTuzes May 15 '21 at 20:32

2 Answers2

4

I'm facing a similar problem and I found a kind of not very ellegant solution. 

binwidth=1
set boxwidth binwidth
bin(x,width)=width*floor(x/width) + binwidth/2.0
set table 'hist.temp'
plot [0:5][0:*] "hist.dat" u (bin($1,binwidth)):(1.0) smooth freq with boxes
unset table

And then you can do the fit of the file as you prefer. I know that probably there are some better way of doing this, but for me it is a fast and working solution. I hope this will be helpful for you.

Cheers!

amaciarey
  • 41
  • 2
1

I used this a nd it worked:

gauss(x)=a/(sqrt(2*pi)sigma)*exp(-(x-mean)**2/(2*sigma**2))

fit gauss(x) 'data.txt' via a,sigma,mean

after 83 iterations GNUplot calculated me a, sigma, and mean

sajad abbasi
  • 1,988
  • 2
  • 22
  • 43
MonaJalilvand
  • 21
  • 3
  • 3
    I think you missed the point: It is not the original data file which Daniel wants to fit a function to, but the data as obtained after using `smooth frequency`. – Christoph Apr 26 '14 at 10:26