Just enter your email
Great resource, thank you. Just working through the MATLAB tutorial raincloudplotstutorial.mlx. The first section that generates codedir, figdir, and datadir has a couple of bugs - it uses double quotes instead of single quotes around the directory names, and a '+' in the square brackets which should be dropped. Cheers, Alex
This is fantastic - thanks for writing it! One idea for simplifying the ggplot2 code: instead of dealing with two different data frames (one summary, one containing the raw data), you could use stat_summary() to create the summary for you. For instance, plot 7 could be generated with:
ggplot(simdat, aes(x=group, y=score, fill = group, colour = group)) +
geom_flat_violin(position = position_nudge(x = .25, y = 0), adjust =2) +
geom_point(position = position_jitter(width = .15), size = .25) +
stat_summary(geom="errorbar", fun.data="mean_cl_boot", position = position_nudge(.25), colour = "BLACK", size = 0.8, width = .1) +
stat_summary(geom="point", fun.data="mean_cl_boot", position = position_nudge(.25), colour = "BLACK") +
ylab('Score') + xlab('Group') + coord_flip() + theme_cowplot() + guides(fill = FALSE, colour = FALSE) +
scale_colour_brewer(palette = "Dark2") +
scale_fill_brewer(palette = "Dark2") +
ggtitle("Figure 7: Raincloud Plot with Mean ± 95% CI")
Great job! I really like the raincloud plot. Perhaps it's easier for users to make raincloud plot if you can convert the R code for making geomflatviolin into an R package, like "raincloud". Otherwise, users have to make a copy of Rrainclouds.R whenever we want to make a new plot in another computer. In addition, is it possible to make raincloud plot as a new type of plot in ggplot2, such as geomraincloud?
Thanks for providing this tutorial. Making it multi-platform was a nice addition.
I believe there is a mistake when the manuscript says 'the philosophy of minimising the “data-ink ratio” (Tufte, 1983)'. Actually, Tufte (1983) advocated: "Maximize the data-ink ratio, within reason." Incidentally, that's what the authors did when presenting only half the probability distribution function for the data.
In the paragraph "To overcome these issues ...", I missed an elaboration on how each of the elements in the raincloud plots address the previously mentioned issues. Currently, this is mostly done in the legend for Figure 4. While the half-split violins and the boxplot or dot-and-whisker seems well justified for me, the datapoints are being justified by watching for striations, which do not appear in the manuscript, and would be less visible with the jittering.
It might be a good idea to elaborate on the merits of the raincloud plot in comparison to the other "Many previous attempts". Can't they be implemented in R, Matlab or Python? But maybe this could be left for another manuscript.