daattali / ggextra

📊 Add marginal histograms to ggplot2, and more ggplot2 enhancements

Hi, I'm Dean 👋

I've been creating widely-used open source projects since 2014. I believe in releasing all my useful software products and knowledge into the wild so that everyone can benefit from them!

My office hours schedule for existing sponsors

You can see a list of my popular tools on my Projects page.

Please consider supporting my work ❤️ for everyone's benefit. I currently spend about half my time on open source and my other half on paying the bills by providing private consulting. My goal is to transition to only open source to reach more people with my work.

ggextra's Issues

Marginal plot not mapped correctly after coord_flip()

I have a geom_boxplot() + coord_flip(). Adding a marginal histogram to the y axis places it on the true x axis. Trying to add it to the x axis throws an error.

Estrange behaviour: ggMarginal (p1, type = "histogram")

Dear @daattali,

I am using ggExtra on a shiny app. I dont know why, but today the rendered plot changed and the histogram is now a "rug" like plot. Any advise on that?

See below a small reproducible example.

library(ggplot2); library(ggExtra)
set.seed(30)
df1 <- data.frame(x = rnorm(500, 50, 10), y = runif(500, 0, 50))
(p1 <- ggplot(df1, aes(x, y)) + geom_point() + theme_bw())

ggMarginal(p1, type = "histogram")
#> Scale for 'x' is already present. Adding another scale for 'x', which
#> will replace the existing scale.

ggMarginal for multi-panel plots

Now it only supports one panel, it will be useful to support multi-panel plots faceted by either facet_wrap or facet_grid

ggplot2 2.2.0 is going to have a lot of breaking changes - fix the code

ggplot2 is being updated again, and it looks like the internal changes in ggplot2 are completely breaking ggMarginal. Again. :(

submit to CRAN

ggMarginal: when main plot is using log/sqrt/etc scale, marginal plots should also

This isn't trivial because I don't know how to generalize looking for the current scale on a plot and using the same scale on another plot. I can manually check "is there a log x/y scale? if so, then add a log x/y scale to the margin plots" and repeat that for every common transformation. But not sure yet how to do it efficiently.

library(ggplot2)
library(ggExtra)
set.seed(123)
df <- data.frame(x = rlnorm(100), y = rlnorm(100))

p1 <- ggplot(df, aes(x, y)) + geom_point() + scale_x_log10() + scale_y_log10()
ggMarginal(p1, type = "histogram")

Error shiny

With the upgrade of shiny says "No layers in plot" which should be the graph:

Refactor ggMarginal and test its internal functions (mostly those that access ggplot2 objects)

Add marginal density/histogram to geom_hex

It will be nice to able to show geom_hex along with two densities on top and right since there are circumstances where scatterplot is too slow due to the huge number of events.

When plotting marginal histograms, the first and last bin are sometimes cut off

As reported by a user in a private email, the first and last bins seem to be cut off. I'm suspecting it's because the first bin's left edge is actually outside of the range of the main plot. For example, if the main scatterplot begins at x = 10, then maybe the first bin is centered at x = 10, but its left edge is at x = 9, which is not within the range.

Support changing the bin size for ggMarginal

As suggested by @_zakahmad on twitter

How to use ggExtra when the X variable is discrete?

I keep getting this error, "Error: StatBin requires a continuous x variable the x variable is discrete. Perhaps you want stat="count"?" How do I solve this issue?

deal with gridExtra brokenness issue

I can't include gridExtra in my package because of a bug that's fixed in their GitHub version but not on CRAN. Get them to submit the fix to CRAN.

baptiste/gridExtra#5

Write an RStudio add-in that lets you select a ggplot object and interactively lets you add marginal plots

Add arguments to 'rotateTextX()'

It would add more convenience to an already convenient function. An idea:

rotateTextX <- function (angle = 90, hjust = angle/90, vjust = 0.5) 
{
  ggplot2::theme(
    axis.text.x = ggplot2::element_text(angle = angle, hjust = hjust, vjust = vjust)
  )
}

ggMarginal() not working with ggplot2.2.1.900

I am trying to replicate the example:

library("ggplot2")
library("ggExtra")

set.seed(30)
df <- data.frame(x = rnorm(500, 50, 10), y = runif(500, 0, 50))
p <- ggplot(df, aes(x, y)) + geom_point()

ggMarginal(p)

Which results the following error:

Error: Don't know how to add o to a plot
In addition: Warning message:
In mean.default(getLimits(pbTop, "y")) :
argument is not numeric or logical: returning NA

---

Session info:

R version 3.3.2 (2016-10-31)
Platform: x86_64-apple-darwin13.4.0 (64-bit)
Running under: macOS Sierra 10.12.3

locale:
[1] en_US.UTF-8/en_US.UTF-8/en_US.UTF-8/C/en_US.UTF-8/en_US.UTF-8

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

other attached packages:
[1] ggExtra_0.6.1      ggplot2_2.2.1.9000

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.9      digest_0.6.12    assertthat_0.1   mime_0.5         grid_3.3.2      
 [6] plyr_1.8.4       R6_2.2.0         xtable_1.8-2     gtable_0.2.0     scales_0.4.1    
[11] lazyeval_0.2.0   miniUI_0.1.1     labeling_0.3     tools_3.3.2      munsell_0.4.3   
[16] shiny_1.0.0.9000 httpuv_1.3.3     colorspace_1.3-2 htmltools_0.3.5  tibble_1.2 

ggMarginal can’t handle Date axes

library(ggExtra)
data(airquality)
airquality$Date <- with(airquality, as.Date(sprintf('1973-%02d-%02d', Month, Day)))

p <- ggplot(airquality, aes(Date, Temp)) + geom_point()
print(ggMarginal(data = airquality, x = 'Date', y = 'Temp'))  # works
print(ggMarginal(p, airquality, 'Date'))                      # doesn’t

the second one throws:

Error: Invalid input: date_trans works with objects of class Date only
Traceback:

1. print(ggMarginal(p, airquality, "Date"))
2. ggMarginal(p, airquality, "Date")
3. ggplot2::ggplot_build(top)
4. lapply(data, scales_transform_df, scales = scales)
5. FUN(X[[i]], ...)
6. unlist(lapply(scales$scales, function(s) s$transform_df(df = df)), 
 .     recursive = FALSE)
7. lapply(scales$scales, function(s) s$transform_df(df = df))
8. FUN(X[[i]], ...)
9. s$transform_df(df = df)
10. f(..., self = self)
11. lapply(df[aesthetics], self$transform)
12. FUN(X[[i]], ...)
13. f(..., self = self)
14. self$trans$transform(x)
15. stop("Invalid input: date_trans works with objects of class Date only", 
  .     call. = FALSE)

split violing plots [feature request]

I love the package, can I suggest a new geom?
Or is there another ggplot extension where this would fit better?

a split violin plot. That is, two densities of groups with the backs to eachother. This stack overflow question has a custum function, but it might be something to add?

https://stackoverflow.com/questions/35717353/split-violin-plot-with-ggplot2

Add support for dotplots

There was a blog post showing how to do many Tufte-style plots in R. The author included ggMarginal but noted that it cannot be used to make dotplots eg

I suppose it could be a nice extra feature to support those kind of plots as well, even though they're very very similar to histograms. I looked into it very briefly and unfortunately the ggplot2::geom_dotplot() doesn't seem to work because when there are many observations the dots go off the screen. There might be a way to somehow dynamically decide how big to make the dots. There can also be other solutions (perhaps useful resource: http://lenkiefer.com/2016/04/06/dot-plots-and-distributions)

If anybody wants to tackle this, feel free to make a PR!

Not default theme set makes grid appear on marginal density

Hi,

When using a theme_set() different from default (for example theme_set(theme_bw())), the marginal densities or histogram use the same theme and thus have grids and lines appearing.

Should ggMarginal suppress warnings when type = boxplot or density

Depending on the type of marginal plot, ggMarginal will pass along one of these warnings:

1. type = histogram - i.e., ggMarginal(data = mtcars, x = "mpg", y = "wt", type = "histogram")

`stat_bin()` using `bins = 30`.
Pick better value with
`binwidth`.

2. type = boxplot - i.e., ggMarginal(data = mtcars, x = "mpg", y = "wt", type = "boxplot")

Warning messages:
1: Continuous x aesthetic -- did you forget aes(group=...)? 
2: Continuous x aesthetic -- did you forget aes(group=...)? 
3: Continuous x aesthetic -- did you forget aes(group=...)? 

3. type = density - i.e., ggMarginal(data = mtcars, x = "mpg", y = "wt", type = "density")

Warning: Ignoring unknown parameters: fill
Warning: Ignoring unknown parameters: fill

In my opinion we should only pass along the warning in case number 1 (i.e., the binwidth warning when type = histogram). The other two warnings are not helpful and are the result of expected behavior. What do you think - should we suppress these warnings?

Histograms axes limits are wrong when using geom_smooth()

Occasionally, I like to have both axes of a scatter have the same range or limits (in the original plot below I set the limits to c(-20,30)). However, to demonstrate the issue, I changed the y-axis limits using scale_y_continuous(limits = c(-40,30)) while having scale_x_continuous(limits = c(-20,30)).

The actual range of the data in the x-axis is: -17.49, 28.56 and on the y-axis is: -9.63, 17.58

I think the location of histograms should adjust and shift when the axes limits are shifted.

support boxplots in ggMarginal

as suggested by a reddit user

Remove external border when combining several graphs

Hi,

I couldn't find the way not to plot the outer frame when combining graphs through ggplot2 + ggExtra + cowplot. I am not sure where I have to tell R, but suspect the issue to lie in ggExtra. Here is an example:
require(ggplot2)
require(cowplot)
require(ggExtra)

Creat a graph

A <- ggplot(mpg, aes(x = cty, y = hwy, colour = factor(cyl))) + geom_point(size = 2.5)

Add marginal histogram

B <- ggExtra::ggMarginal(A,type = 'histogram', margins = 'x', size = 9)

Combine through cowplot

combo <- plot_grid(B,B,labels=c("A","B"))
plot(combo) # looks fine

Re-combine through cowplot

plot_grid(B,combo,ncol=1,rel_heights = c(2,3))

The question is also posted on Stackoverflow, here.

Thanks for any help!

Initial questions

Hi there -

Two initial questions I have come across while refactoring:

1. Are we sure that this is needed for ggplot2 V 2.2 and above (lines 96-99 of ggMarginal.R):

 # after ggplot2 v1.0.1, layers became strict about parameters
  if (type == "density") {
    extraParams[['fill']] <- NULL
}

I'm getting the behavior that I would want (a warning from ggplot: "Warning: Ignoring unknown parameters: fill") when I drop those lines from ggMarginal and I run this code:

p <- ggplot2::ggplot(mtcars, ggplot2::aes(wt, mpg)) + ggplot2::geom_point()
ggExtra::ggMarginal(p, type = "density", fill = "black")

I also don't understand why this line would be needed for a density geom and not boxplot/histogram but that's probably just my ignorance.

2. Regarding this bit of code (lines 113-143 of ggMarginal.R):

if (missing(p)) {
    if (missing(data) || missing(x) || missing(y)) {
      stop("`data`, `x`, and `y` must be provided if `p` is not provided",
           call. = FALSE)
    }
    p <- ggplot2::ggplot(data, ggplot2::aes_string(x, y)) + ggplot2::geom_point()
    x <- as.symbol(x)
    y <- as.symbol(y)
  } else {
    if (missing(data)) {
      if (methods::is(p$data, "waiver")) {
        stop("`data` must be provided if it is not part of the main ggplot object",
             call. = FALSE)
      }
      data <- p$data
    }
    if (length(p$mapping) == 0) p$mapping <- p$layers[[1]]$mapping
    if (margins != "y" && missing(x)) {
      if (is.null(p$mapping$x)) {
        stop("`x` must be provided if it is not an aesthetic of the main ggplot object",
             call. = FALSE)
      }
      x <- p$mapping$x
    }
    if (margins != "x" && missing(y)) {
      if (is.null(p$mapping$y)) {
        stop("`y` must be provided if it is not an aesthetic of the main ggplot object",
             call. = FALSE)
      }
      y <- p$mapping$y
    }
  }

...We already make sure in the if statement that the user provided data, x, and y if they didn't provide p. Under what circumstance could they pass this test and be thrown an error in the else block? In other words, under what circumstances would the user supply a ggplot object p but you still need them to supply x, y, or data? It looks like it may be an assertion that you ggplot still keeps x, y, and data where you think they will be in the plot object, but it's hard to tell if you're testing that or user params.

ggExtra with facets

Is it possible to use your fantastic package to create a plot described in this question :)?

add travisCI

add demo shiny app to make it easy to experiment before installing

On smaller monitors, results plot is "cut-off" at bottom x-axis.

NOTE: Problem only happens on smaller monitors.
My 2nd monitor, where I run R studio, is 14 inches diag.
(No problem on larger monitors)

Using your great mtcars ggExtra Margins example,
in the latest Rstudio / Linux.

From the Console panel,
this works ok:

library(ggplot2)
library(ggExtra)
myplot <- ggplot(mtcars, aes(wt, mpg)) + geom_point()

But...using the Rstudio "addin"
from the Code panel,
shows:
http://imgur.com/epvDELb

See?
The bottom part of the plot is "cut-off" (below mpg=10)
...can not see the wt x-axis.

So yes, everybody with smaller monitors
may have that problem.

Q: (possible solution?)
Is there a way
to make the results Margins popup window,
"user-resizable"?.

I'm sure that if I could use the mouse
to "resize" that popup,
the hidden parts will re-appear OK,
even on smaller monitors.

btw:
There is lots of unused screen space
around that popup,
so allowing the user to mouse-resize it,
would certainly solve the problem! :-)

Improve ggMarginal ... param to allow specifying diff params to x and y margins

look into using NSE in the demo gadget

Adding marginal plots for grouped data

Hi Dean,

Thank you for your work in ggExtra package, which makes it really easy to add marginal plots to ggplots.

It would be highly appreciated, if one can add marginal plots for grouped data as illustrated here and here

Suggestion: Improve ggMarginal()so that it reacts to the mapping arguments.

sp <- ggplot(iris, aes(Sepal.Length, Sepal.Width))+
  geom_point(aes(color = Species))

ggMarginal(sp, data = iris, mapping = aes(color = Species) )

Best regards,

horizontal geoms

would you be interested in a contribution (pull request) of some horizontal variants of geoms such as linerange (useful when trying to use faceting and flipped coordinates together, which is not currently feasible in ggplot2)?

e.g. see:

• tidyverse/ggplot2#1128
• tidyverse/ggplot2#1261

Error running ggMarginal

First, thank you very much for developing such a nice package!

I am having this error when running the function 'ggMarginal'

Error in data.frame(x = .Primitive("log10"), PANEL = c(1L, 1L, 1L, 1L, :
arguments imply differing number of rows

Any idea how to solve the problem?

Maybe this is related to the log10 in the formula in ggplot2.

The code used:

p1 <- ggplot(dat.crop, aes(y = log10(AGr), x = log10(herb + 1))) +
geom_point(size = 3, color = "tomato")+
geom_smooth(method = "lm")
ggMarginal(p1, type = "histogram")

Thanks in advance

legends offsets histograms

Hi,
when you have legends, the marginal plots are not displayed correctly to the axis. See the attached files for an example.
Regards,
Yadhu

Adding y-axis labels on the marginal plots

Hi,

I would like to add y-axis labels for the marginal plots. Is there a way to do it using ggMarginal command?

Thanks,
Lavanya

Add visual tests for ggMarginal (all versions of ggplot2 >= 2.2.0)

Add visual tests for ggMarginal (current version of ggplot2)

Adding axis to marginal denstity plot and histograms

Great package, anyway I think it could be useful to have the possibility to show the axis of the density plot/histogram, in order to be able to compare distributions of subsets of data (see pict):

As you can see the maximum of each the density distribution is the same, therefore it's not possible to appreciate the differences in density value.
Thanks in advance.
Sebastiano

ggMarginal should carry theme of title through

Right now ggMarginal only carries the color and font face of the scatter plot's title through to the final plot. The text grob that we use for a title in the final plot is created here:

    title <- grid::textGrob(
      pb$plot$labels$title,
      gp = grid::gpar(col = pb$plot$theme$plot.title$colour,
                      fontsize = 16, fontface = pb$plot$theme$plot.title$face)
    )

It's easy to pull out the ggplot2 titleGrob that contains all of the theme elements desired by the user, and drop that grob in instead of using grid::textGrob. I will include this in the next PR.

Support for marginal violin plots

Histogram margin should be inside of legend

Minimal example:

library("ggExtra")
# Create some data
set.seed(1234)
x <- c(rnorm(500, mean = -1), rnorm(500, mean = 1.5))
y <- c(rnorm(500, mean = 1), rnorm(500, mean = 1.7))
col <- c(rnorm(500, mean = 1), rnorm(500, mean = 1.3))
df3 <- data.frame(x, y, col)
# Scatter plot of x and y variables and color by groups
sp2 <- ggplot(df3,aes(x, y, color=col)) + geom_point()
ggMarginal(sp2 + theme_gray(), type="histogram")

The legend and the y-axis histogram should be switched.

Printing a ggMarginal object to a non interactive device results in a leading blank page

This is fixed by @crew102 's PR #47

Hi. I am having a lot of trouble getting this to work. First, the grid package isn't available for the R version I have (latest) and then even though the package is loaded, I get a msg saying to '...please install gridExtra' What am I missing?

Hi. I am having a lot of trouble installing gridExtra. First, grid doesn't install under the current version of R and then when I try running running the package I get "Error: gridExtra package is required for this function to work. Please install it." even though the package is installed. Help?

X-axis label too close to edge of plot - need a bit of margin because some letters get cut off

At some point in the past year, the bottom margin of the x axis label has disappeared. The label is stuck to the bottom of the image without any space, and any letters that go lower the the standard line (such as "g" "j" "p" "q" "y") are cut off.

I'm not sure if this is because of newer ggplot version, or because of the refactoring from #26

ggMarginal: if main plot has a title, add the title on top of the top margin

`Error: no layers in the plot` still occures for gridExtra_0.9.1

Error: no layers in the plot still occures when you have an old version of gridExtra (it does not work for me for version gridExtra_0.9.1) so you should ensure that user has at least version 2.0.0 of gridExtra. Can you fix this somehow or add it to the vignette/info/description?

Add webhook to trigger build when new version of ggplot2 passes

ggMarginal(p) invalidates the boldness of the main plot title

ggMargin is great. But it changes bold main title to plain text, which is not desired.

p <- ggplot(mtcars, aes(x=wt, y=mpg)) + geom_point() + theme_bw() +
labs(title="Main Title", y="MPG") +
theme(plot.title=element_text(face="bold", size=14),
axis.title.y=element_text(face="bold", size=14))

p's main title and y are bold.

ggMarginal(p, margins="y", size=4)

Now main title is changed to plain text. The y title font face is not affected. Can you maintain the font face of main title?

Notes suggested by @baptiste

as suggested by @baptiste

1. See if the hacks caused by gridExtra not playing nicely in packages can be removed
2. add a note about the fact that this is not the same ggExtra as the one from many years ago