The latest release of vioplot (it may also happen
in earlier releases), has an irritating typo:

Line 333 of vioplot.R contains a spurious:
print(ylim)

which generates output if a value is assigned to this argument.

Reported by @Derek-Jones

vioplot(stuff, cex =0.9) does not have the desired effect, while

par(cex=0.9)
vioplot(stuff)

works fine.

Will report issues here, now I know about the github repro (it's not listed on CRAN)

Trying to plot two parallel violin plots using the formula interface. Here's a minimal example and the (surprising) outcome:

library("vioplot")
DF <- data.frame(x=c(rnorm(500),  rnorm(250, mean=-1, sd=.6), rnorm(250, mean=1, sd=.6)), group=factor(rep(1:2, times=c(500, 500))))
vioplot(x ~ group, data=DF)

which produces this result

I'm fairly sure the problem is caused by the last line of vioplot.formula

expression(vioplot( 1, 2 , xlab = xlab, ylab = ylab, names = names, ...))

Since the categories are named the numbers 1 and 2, they are parsed as numbers by vioplot.default and not as the labels referencing the data in the datas column. As far as I can see the default is not easily modified since the first two arguments are also used as the actual labels in the plot.

Looking at the code, the parameter cex.names is assigned to, but never passed to anything, to have an effect (apart from testing whether it has a non-default value).

The xaxt argument is ignored in calls to vioplot, e.g., xaxt="n" has no effect.

Internally, this parameter does not appear in calls to plot.window.

The vioplot function provides an error if you are using a data frame with levels, that includes a space. Setting the names argument does also not help. The only thing that helps is changing the levels of the factor variable.
A minimal example is attached.
Error_spaces.txt

Thanks for the awesome package. The following is the issue I am facing:

When giving the vioplot too many of the same value (and only that value) it produces an error.

These work fine:

> vioplot(x = c(0))
> vioplot(x = c(rep.int(0, 100)))

Adding one order of magnitude and it breaks:

vioplot(x = c(rep.int(0, 1000)))
vioplot(x = c(rep.int(1, 1000)))

The error I get is:

Error in cut.default(x, breaks = breaks) : 'breaks' are not unique

To clarify, if there is even one more number with another value, it will work as intended:

vioplot(x = c(rep.int(c(1), 10000), 2))

My session's info:

R version 4.0.3 (2020-10-10)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Manjaro Linux

Matrix products: default
BLAS:   /usr/lib/libopenblasp-r0.3.12.so
LAPACK: /usr/lib/liblapack.so.3.9.0

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C              
 [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8    
 [5] LC_MONETARY=fi_FI.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=fi_FI.UTF-8       LC_NAME=C                 
 [9] LC_ADDRESS=C               LC_TELEPHONE=C            
[11] LC_MEASUREMENT=fi_FI.UTF-8 LC_IDENTIFICATION=C       

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods  
[7] base     

other attached packages:
[1] vioplot_0.3.5   zoo_1.8-8       sm_2.2-5.6     
[4] beeswarm_0.2.3  varhandle_2.0.5 RCurl_1.98-1.2 
[7] XML_3.99-0.5   

loaded via a namespace (and not attached):
 [1] rstudioapi_0.13 knitr_1.30      magrittr_1.5   
 [4] hms_0.5.3       lattice_0.20-41 R6_2.5.0       
 [7] rlang_0.4.8     highr_0.8       tcltk_4.0.3    
[10] tools_4.0.3     grid_4.0.3      xfun_0.19      
[13] ellipsis_0.3.1  yaml_2.2.1      tibble_3.0.4   
[16] lifecycle_0.2.0 crayon_1.3.4    readr_1.4.0    
[19] vctrs_0.3.4     bitops_1.0-6    compiler_4.0.3 
[22] pillar_1.4.6    pkgconfig_2.0.3

It seems the arguments las and ylab are not implemented in the package. I do not get any error/warning but they do not work either.

I'm just reporting that as a fresh user of vioplot, I initially got confused by the way the h parameter of vioplot() is documented.

I was looking for the density bandwith tuning argument, but the doc says h is the "the height for the density estimator, if omit as explained in sm.density, h will be set to an optimum". I interpreted this as a tuning for the height of the density plot (which is in fact the width when the violin is drawn vertically), but this happens to be tuned by wec (however, to add to the confusion, the code of vioplot then computes in internal variable "hscale" based on wec...).

My proposition is to use a documentation of the h parameter based on the doc of sm.density ("a vector of length one, two or three, defining the smoothing parameter. A normal kernel function is used and h is its standard deviation. If this parameter is omitted, a normal optimal smoothing parameter is used."). The important keywords to appear for me are "density smoothing parameter".

References

Places where "height" is used:
https://github.com/TomKellyGenetics/vioplot/blob/master/R/vioplot.R#L15
https://github.com/TomKellyGenetics/vioplot/blob/master/man/vioplot.Rd#L181

Doc of vioplot (0.3.7):
https://www.rdocumentation.org/packages/vioplot/versions/0.3.7/topics/vioplot

Doc of sm.density:
https://www.rdocumentation.org/packages/sm/versions/2.2-5.4/topics/sm.density

Thank you for a great package.

For some cases, it would be helpful to use a histogram rather than a kernel density. The ability to use R's hist function and breaks argument would be very easy to use. This functionality is also not currently available via ggplot.

Somebody new to vioplot might pass an xlim argument (this this is available in many other plot functions; I did this in one of my plots).

vioplot(sample(44)) # works as expected

vioplot(sample(44), xlim=c(1, 2)) # surprising behavior

The xlim causes a second plot to appear, next the the sample violin plot.

There is no option to change the color of the center line, running down the middle of the violin.

If you don't want to add another option, then perhaps make it the same as the border color?

When col is a vector, its values are not reused (the online help says they are).

Looking at the code, col[i] should be col[1+(i-1)%length(col)]

The other color parameters have the same problem, e.g., border

I am using R version 4.0.4 (2021-02-15) and vioplot_0.3.7.

vioplot(outcome ~ intervention, data = d, xaxt = 'n') works as expected: intervention values are not marked on the x-axis, but outcome values are marked on the y-axis.

vioplot(outcome ~ intervention, data = d, xaxt = 'n', horizontal = TRUE) also does something understandable: intervention values are not marked on the (vertical) "x-axis", but outcome values are marked on the (horizontal) "y-axis." This is not what the current version of boxplot does, but it's possible to work around that.

But both vioplot(outcome ~ intervention, data = d, yaxt = 'n') and vioplot(outcome ~ intervention, data = d, yaxt = 'n', horizontal = TRUE) result in neither intervention values nor outcome values being marked on their respective axes. This seems like it has to be wrong.

I took a look at the code, and this clearly is at least partially a reflection of the fact that all the if(xaxt !="n") conditionals in vioplot.R are wrapped inside if(yaxt !="n") conditionals. However, when I added else clauses for the latter, with similar code for handling the x-axes alone inside it, this made the two snippets above work as I expected them to, but still resulted in this code:

par(yaxt = 'n')
vioplot(outcome ~ intervention, data = d)

suppressing the plotting of both axes. I do not know why this is.

vioplot(sample(40), xlog=TRUE, horizontal=TRUE)
vioplot(sample(40))
Error in plot.window(xlim, ylim, log = log, asp = asp, bty = bty, cex = cex, :
"log=" specification must be character

A vioplot(sample(40), ylog=FALSE) makes the message go away

I've received several related questions which can be resolved by overlaying base R graphics and integrating vioplot with other plotting functions.

"vioplot()" functions similar to "plot()" and passes input arguments from "par()".

For example it is possible to add additional annotations. As requested by email here is an example:

# generate dummy data
a <- rnorm(25, 3, 0.5)
b <- rnorm(25, 2, 1.0)
c <- rnorm(25, 2.75, 0.25)
d <- rnorm(25, 3.15, 0.375)
e <- rnorm(25, 1, 0.25)
datamat <- cbind(a, b, c, d, e)
dim(datamat)
#> [1] 25  5
# violin plot
library("vioplot")
#> Loading required package: sm
#> Warning: package 'sm' was built under R version 4.1.3
#> Package 'sm', version 2.2-5.7: type help(sm) for summary information
#> Loading required package: zoo
#> Warning: package 'zoo' was built under R version 4.1.3
#> 
#> Attaching package: 'zoo'
#> The following objects are masked from 'package:base':
#> 
#>     as.Date, as.Date.numeric
vioplot(datamat, ylim = c(0, 5))
# compute medians
data.med <- apply(datamat, 2, median)
data.med
#>         a         b         c         d         e 
#> 3.0475551 2.2365210 2.7504934 3.0730919 0.9803336
#overlay medians
lines(data.med, lty = 2, lwd = 1.5)
points(data.med, pch = 19, col = "red", cex = 2.25)

^{Created on 2022-11-07 with reprex v2.0.2</sup](https://reprex.tidyverse.org%29%3C/sup)>}

It is also possible to modify the axes labels and titles as shown in issue #16

library("vioplot")
#> Loading required package: sm
#> Warning: package 'sm' was built under R version 4.1.3
#> Package 'sm', version 2.2-5.7: type help(sm) for summary information
#> Loading required package: zoo
#> Warning: package 'zoo' was built under R version 4.1.3
#> 
#> Attaching package: 'zoo'
#> The following objects are masked from 'package:base':
#> 
#>     as.Date, as.Date.numeric
outcome <- c(rnorm(25, 3, 1), rnorm(25, 2, 0.5))
intervention <- c(rep("treatment", 25), rep("control", 25))
table(intervention)
#> intervention
#>   control treatment 
#>        25        25
names(table(intervention))
#> [1] "control"   "treatment"
unique(sort(intervention))
#> [1] "control"   "treatment"
intervention <- as.factor(intervention)
levels(intervention)
#> [1] "control"   "treatment"
d <- data.frame(outcome, intervention)
vioplot(outcome ~ intervention, data = d, xaxt = 'n', yaxt = 'n', 
        main = "", xlab = "", ylab = "")
axis(side = 1, at = 1:length(levels(intervention)), labels = levels(intervention))
mtext("custom x labels for intervention", side = 1)
mtext("custom y labels for outcome", side = 2)
title(main = "example with custom title", sub = "subtitles are supported")

^{Created on 2022-11-09 with reprex v2.0.2}

I'll share these examples publicly for future reference. This issue indicates these are under consideration in updated documentation or vignettes.

1. The titles or names of the axis are not properly overwritten when I use the code
   title(xlab = "Species in Iris table", ylab = "Sepal Length in iris table"), but the original names of the axes are still there! What can I do to have them properly removed.

2. Sometimes it would be very helpful to have the legend OUTSIDE the actual plotting area, e.g. under the x-axis or right to the plotting area. I see only options where it is inside it. If you have other data than the example from iris, the legend can hide (part of) the violins, which is not nice. In the example I put the legend on the left bottom to show what I mean. You don't see the first violin now.

How to have the legend OUTSIDE the polling area?

Here is the code that produced the plot below.:
`
library(vioplot)

iris_large <- iris[iris$Sepal.Width > mean(iris$Sepal.Width), ]
iris_small <- iris[iris$Sepal.Width <= mean(iris$Sepal.Width), ]

vioplot(Sepal.Length~Species, data=iris_large,
col = "palevioletred",
plotCentre = "line", side = "right")

vioplot(Sepal.Length~Species, data=iris_small,
col = "lightblue",
plotCentre = "line", side = "left", add = T)

title(xlab = "Species in Iris table", ylab = "Sepal Length in iris table")

legend("bottomleft", fill = c("lightblue", "palevioletred"),
legend = c("small according to iris table", "large according to iris table"),
title = "Sepal Width as classified in iris table")
`

To compare multiple groups of histogram densities, it helps to adjust the wex. Here is a MWE that does so, which you may want to add to a vignette.

dlist1 <- lapply(c(10,20,30,40), function(n) runif(n))
dlist2 <- lapply(c(100,200,300,400), function(n) runif(n))

hscale1 <- sapply(dlist1, function(r){
    max(hist(r, plot=F, breaks=seq(0,1,by=.05))$density)})
histoplot(dlist1, side='left', col=grey(.3),
    breaks=seq(0,1,by=.05), add=F, pchMed=NA, drawRect=F, border=NA,
    wex=hscale1/length(hscale1))

hscale2 <- sapply(dlist2, function(r){
    max(hist(r, plot=F, breaks=seq(0,1,by=.05))$density)})
histoplot(dlist2, side='right', col=grey(.7),
    breaks=seq(0,1,by=.05), add=T, pchMed=NA, drawRect=F, border=NA,
    wex=hscale2/length(hscale2))

Sometimes, it is helpful to see the raw counts instead.

dvec <- length(unlist(c(dlist1, dlist2)))/4

histoplot(dlist1, side='left', col=grey(.3),
    breaks=seq(0,1,by=.05), add=F, pchMed=NA, drawRect=F, border=NA,
    wex=sapply(dlist1, length)/dvec*hscale1/length(hscale1))
histoplot(dlist2, side='right', col=grey(.7),
    breaks=seq(0,1,by=.05), add=T, pchMed=NA, drawRect=F, border=NA,
    wex=sapply(dlist2, length)/dvec*hscale2/length(hscale2))

It may also benefit some users to pass density and angle arguments to the histograms (ultimately rect) and create outer legends

hist(runif(100), density=c(10,20), angle=c(22,90+22) ,col=1)

outer_legend <- function(...) {
  opar <- par(fig=c(0, 1, 0, 1), oma=c(0, 0, 0, 0), mar=c(0, 0, 0, 0), new=T)
  on.exit(par(opar))
  plot(0, 0, type='n', bty='n', xaxt='n', yaxt='n')
  legend(...)
}
outer_legend('topright', pch=15, density=c(10,20), angle=c(22,90+22), col=0, legend=c('Y','N'))