I have successfully made a nice looking histogram in gnuplot with a normal scale but when I add
set logscale y
to my code, the data disappears.
I discovered that logscale does not work with the function I am using to make the histogram (shown in the following link). https://stackoverflow.com/a/2538846/2506689
Does anyone have any suggestions on how to fix this?
set table to the rescue!
Note that I typed all of this into the interactive prompt and copy/pasted my terminal contents. As such, my commands that follow are prefixed by a gnuplot> which you won't include in your script :-)
First, I generate some data using python ...
import numpy as np
np.savetxt('foobar.txt',np.random.random(1000))
That wasn't hard. Now time to set up gnuplot functions/constants:
gnuplot> binwidth = 0.05
gnuplot> bin(x,width)=width*floor(x/width)
gnuplot> plot 'foobar.txt' using (bin($1,binwidth)):(1.0) smooth freq with boxes
Ok, it works with non-logscale data. That's good. Lets write that data into a separate file using set table
gnuplot> set table 'foobar.table'
gnuplot> plot 'foobar.txt' using (bin($1,binwidth)):(1.0) smooth freq with boxes
gnuplot> unset table
Now I look at the data gnuplot wrote out to see what's there.
gnuplot> !head foobar.table
# Curve 0 of 1, 21 points
# Curve title: "'foobar.txt' using (bin($1,binwidth)):(1.0)"
# x y xlow xhigh type
0 40 0 0 i
0.05 57 0.05 0.05 i
0.1 52 0.1 0.1 i
0.15 56 0.15 0.15 i
0.2 49 0.2 0.2 i
0.25 55 0.25 0.25 i
Unfortunately, it looks like xlow and xhigh are always the same (possible bug?). But that's Ok, we're using a constant binwidth anyway. We'll just use that as the width.
gnuplot> set logscale y
gnuplot> plot 'foobar.table' u 1:2:(binwidth) w boxes
I should note that I've been a little loose with my box positions. To really get it right, you probably need to shift the center of the boxes to the right by half a binwidth:
gnuplot> plot 'foobar.table' u ($1+0.5*binwidth):2:(binwidth) w boxes