I have a data file with two columns: Xi and Yi. I'd like to plot Xi vs. (Yi-1 - Yi)/Yi-1 for i>1. Is is possible to do that in GNUPlot direclty?
Add a comment
|
2 Answers
Yes it is possible with gnuplot directly -- It's just not easy:
firstval = NaN
yi1(yi) = (returnval=firstval, firstval=yi, returnval)
plot "datafile" using 1:((yi1($2)-$2)/returnval)
You need to use inline functions. inline functions are of the form:
funcname(args,...) = (statement1,statement2,...,statementn, return_value)
Here I just created a function to hold the last value it was passed. Unforunately, this solution gets a little more ugly since I couldn't call yi1 twice in the using specification (the second time, I would get the wrong return value), so I had to reference the variable holding the return value directly. It's not pretty, but it works. You could probably "pretty" it up a little bit by passing $0 (the line number) and only updating when $0 changes, but it's probably not worth it for this hack.
Comments
To do this in gnuplot directly is tricky. The problem is that you have to use the (i-1)th element in your calculation, which gnuplot can't do automatically. It can do simple calculations, but only on the same row, for example something like
plot "datafile" using ($1):(($2-$1)/$2)
would be easy.
For what you neet to do I'd recommend octave, or you could prepare your data file using a spreadsheet application.
In octave, you could plot this like:
d = load("datafile")
plot(d(2:end, 1), d(1:end-1, 2)-d(2:end, 2)./d(1:end-1, 2))
3 Comments
mgilson
You can do this sort of thing if you want to with only gnuplot (see my solution), but it quickly becomes cumbersome if you want the i-2, or i-3 element. And i+1 and i+2 elements cannot be retrieved (but that's not too much of a problem -- just shift your indexing :).
mata
@mgilson - didn't think about such a solution :) the only problem I see is that you can't do a
replot without reseting firstvalmgilson
Definitely true -- Although I don't really find myself replotting very often these days.
