Gnuplot is a power, open-source, plotting program that I have been using since undergraduate school. It is simple enough to plot data in just a few lines, flexible enough to produce publication-quality figures, and power enough to analyze real scientific data sets.

Gnuplot is a command line program. This makes it very powerful because it can be scripted, but it also makes it difficult for new users to learn, especially if they do not have experience using the command line.

Installing Gnuplot on Linux is simple. You just do whatever the equivalent of `apt install gnuplot` is on your distribution. Gnuplot also runs on

I wanted to require my intro-level physics classes to use Gnuplot for some simple data analysis, but I didn't want to deal with issues related to installing it. What if someone was using a Chromebook for example? I found an online gnuplot here, but it uses Gnuplot 4 and I've been using Gnuplot 5 for quite a while. So I decided to learn how build WebAssembly apps and create my own version for students to use.

If you want to have students use Gnuplot in lab, feel free to let them use this site. Or, you can build a version and host it on your own server. The code is on Github.

One of the great features of Gnuplot is how easy it is to fit a function to data. In a few lines of code, you can fit an arbitrary function to a data set. Gnuplot's nonlinear least-squared fit function uses the Marquardt-Levenberg algorithm.

Here again, Gnuplot makes it simple to do a basic fit (select "Linear Regression" at the top of the page and click "Load" to see an example), but provides many powerful options. For example, Gnuplot can handle uncertainty in both the x and y data. These uncertainties are factors into the fit, and affect both the values and error reported by Gnuplot for the fit parameters.

Even though Gnuplot's fit function is easy to use, it's also quite good. NIST (the National Institute of Standards and Technology) maintains a set of data sets that can be used to test nonlinear regression software. I ran these tests through Gnuplot, and it produced the expected results to within a few percent for almost all examples. Only a few tests failed, and a few more gave differed significantly from the expected values. The several of the tests were designed to be difficult, so this isn't surprising. To see a description of the tests, visit the NIST website. Thre results for each test are listed in the table below (you should be able to click on the table headers to sort each column).

The videos below are simple how-to videos for doing some basic tasks in gnuplot. If your interested in more sophisticated stuff, have a look at the Gnuplot online documentation, one of the many online tutorials, or sign up to take my Scientific Computing and Productivity course.