You must log in or register to comment.
Why not just use gnuplot?
I only know of gnuplot, how does ggplot differs from gnuplot?
I like Makie #JuliaGang
There are dozens of us #JuliaGang
Oh this touches close to him. I got into pgfplots since it would generate plots in latex at compile time and keep fonts consistent, etc. plots looked amazing though.
The worst was when a colleague couldn’t get a pdf to upload into a google doc, so he just made an ugly ass bar chart in excel for the final draft since that was easier. The only reason he could do that so quickly was because he could read the data so easily from the plot I made. Ugh. Still burns
R with the tidyverse package is amazing once you get over the learning curve. It’s so much easier to simply type a few lines of code then to fiddle with the Excel GUI, plus the ability to customize the plot is much, much better in R.
Yes making a simple plot in Excel is relatively easy, but try making something evening remotely complex and it’s terrible. A box plot is a great example of this, 2 lines of plotting code in R for a basic plot but an absolute nightmare to create in Excel.
Afaik, you can’t even make a box plot on excel without 3rd party plugins. You can maybe hack together one with a bar chart, and hiding/narrowing parts of it so that it sort of resembles a box plot if you never ever open the data.
Why not just cut to the chase and do it in Libre Office Calc?
Hot tip - export a basic plot to svg and format it in Inkscape. OBVIOUSLY DO NOT CHANGE ANYTHING THAT MODIFIES THE DATA OR RESULTS but it’s much easier to get a consistent look and feel in Inkscape than ggplot
Hotter tip - just make it all in inkscape, no one fact-checks anyway!
There’s inkscape plugin to make barplots and piecharts. Why don’t we add for more? Honestly we could even make it just take r code or python code.
both foss software, right?
Matplotlib and seaborn are my Go-to
Valid. I remember being very frustrated with ggplot’s documentation, only to realize they HARD CODED a plot parameter that could not be accessed. Matplotlib + Seaborn works fantastic, and I can do whatever I want without being babied by ggplot
Matplotlib is still a pain in the ass when you need to customize something seaborn can’t handle
Matplotlib + export to SVG + final adjustments in Inkscape is my pick.
Use plotnine if you want a ggplot-like API. Matplotlib is too low level for most stuff so it needs an abstraction layer such as plotnine or seaborn. Also, I understand the original devs desire to copy Matlab plotting standards, but that language is so bad.
I rather use the r-base
plotfunctions for everything in R. It is more flexible especially when you need to tweak the visuals.It can also be EXTREMELY long code to do something relatively simple. I bounce between base and ggplot, and use ggh4x for some oddly difficult stuff in ggplot.
Honest question: do you think this could improve with practice? Or does the ggplot workflow necessarily makes it all slower?
Once you have figured it out, it’s actually a nice workflow. Don’t get me wrong, when I’m not publishing a paper, I quickly forget all commands, my whole setup etc. and start from scratch, cursing a lot and retracing my steps in the history, basically re-learning the framework. I’d still never move away from ggplot2.
It absolutely improves with practice, and once you have settled on an aesthetic you like you can simply reuse the code, e.g. store all your color/line properties in a variable and just update each figure with that variable
My thesis had something like 30 figures, and at multiple points I had to do things like “put these all on a log scale instead” or “whoops, data on row 143,827 looks like it was transcribed wrong, need to fix it”
While setting everything up in ggplot took a couple hours, making those changes to 30 figures in ggplot took seconds, whereas it would have taken a monumental amount of time to do manually in excel
graphpad is my #1
