Boxplot outliers are shown in black using ggplotly

[dfrail24]

If I create a boxplot in ggplot2 and convert it using ggplotly command, the outliers are outlined in black.
Here is a simple example:

library(ggplot2)
library(plotly)

p <- ggplot(mpg, aes(class, hwy))
g <- p + geom_boxplot(aes(colour = "red"))
ggplotly(g)

ggplot would show this chart:

whereas plotly would show this chart:

Is this something that can be fixed?

[jonocarroll]

This persists even when the outliers should be discarded, in the examples also

library(plotly)
set.seed(123)

df <- diamonds[sample(1:nrow(diamonds), size = 1000),]

p <- ggplot(df, aes(cut, price, fill = cut)) + 
  geom_boxplot(outlier.shape = NA) + 
  ggtitle("Ignore outliers in ggplot2")

# Need to modify the plotly object and make outlier points have opacity equal to 0
p <- plotly_build(p)

p$data <- lapply(p$data, FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

p

[petehilljnr]

I managed to set the opacity property of the outliers using the code below. This seems to work for the faceted charts I have tried so far also.

library(ggplot2)
library(plotly)

set.seed(123)

df <- diamonds[sample(1:nrow(diamonds), size = 1000),]

p <- ggplot(df, aes(cut, price, fill = cut)) + 
  geom_boxplot(outlier.shape = NA) + 
  ggtitle("Ignore outliers in ggplot2")

# Need to modify the plotly object and make outlier points have opacity equal to 0
p <- plotly_build(p)

for(i in 1:length(p$x$data)) {
  p$x$data[[i]]$marker$opacity = 0
}

p

[jonocarroll]

The replacement lapply code is then

p$x$data <- lapply(p$x$data, FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

(note p$x$data rather than p$data). I'm happy to PR this to the documentation if someone can point to the source.

[dmattek]

The problem is that when you also have geom_jitter in the plot (in addition to geom_boxplot), the lapply part will remove all the points. Is there a way to selectively remove outliers that belong to geom_boxplot only?

[ningjingzhiyuan507]

p$x$data <- lapply(p$x$data, FUN = function(x){ x$marker$line$width = 0 return(x) })

modify marker$line$color

[brshallo]

The problem is that when you also have geom_jitter in the plot (in addition to geom_boxplot), the lapply part will remove all the points. Is there a way to selectively remove outliers that belong to geom_boxplot only?

You can use the code above and just index to the layer you want to remove, e.g. say the boxplot outliers are on the first layer.

p$x$data[1] <- lapply(p$x$data[1], FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

[kojisposts]

Hi! Just wanted to bring this issue to your attention again, as none of the workarounds mentioned above seem to be working (and aren't working in the documentation either)!
(There's also an interesting phenomenon where, for coloured barplots, the most extreme outliers are coloured with black outlines, but closer to the barplot, they're black with coloured outlines, i.e. the reverse.)

[cpsievert]

There's a WIP here #1514 that fixes this issue, feel free to test it out and let me know if you run into problems.

[jcoronel25]

I didn't see the solution being mentioned #1514 on the last release. I was to get the visual I wanted by altering the lapply function to filter only layer that are type == "box"

p$x$data <- lapply(p$x$data, FUN = function(x){

  if (x$type == "box") {
    x$marker = list(opacity = 0)
  }
  return(x)
})

[isaaczhao23]

This will do the trick for the original question coloring outliers! Plotly differentiates outliers from extreme outliers. We go under the hood and override all outlier colors manually.

library(ggplot2)
library(plotly)

p <- ggplot(mpg, aes(class, hwy)) + geom_boxplot(color="red")
output = ggplotly(p)

# overrides black outline of outliers
output$x$data[[1]]$marker$line$color = "red"
# overrides black extreme outlier color
output$x$data[[1]]$marker$outliercolor = "red"
# overrides black not as extreme outlier color
output$x$data[[1]]$marker$color = "red"

output

[meldarionqeusse]

Using this code:

set.seed(123)

df <- diamonds[sample(1:nrow(diamonds), size = 1000),]
df <- subset(df,df$cut=="Fair")


p <- ggplot(df, aes(cut, price, fill = cut)) + 
  geom_boxplot() 

# Need to modify the plotly object and make outlier points have opacity equal to 0
fig <- plotly_build(p)

fig$x$data <- lapply(fig$x$data, FUN = function(x){
  x$marker$line = list(opacity = 0)
  x$marker$line$color = list(opacity = 0)
  x$marker$outliercolor = list(opacity = 0)
  x$marker$color = list(opacity = 0)
  return(x)
})
fig

I was able to get to this plot:

As you can see the outliers are different. I am aiming to get a transparent outlier in this case but the extreme outlier remains filled. From the output of fig$x$data I cannot understand which parameter effects the extreme outlier. Looks at the html styling output I can say that the two points are definitely getting different rgb values and fill-opacity values.

[[1]]$marker
[[1]]$marker$opacity
[1] NA

[[1]]$marker$outliercolor
[[1]]$marker$outliercolor$opacity
[1] 0


[[1]]$marker$line
[[1]]$marker$line$opacity
[1] 0

[[1]]$marker$line$color
[[1]]$marker$line$color$opacity
[1] 0



[[1]]$marker$size
[1] 5.669291

[[1]]$marker$color
[[1]]$marker$color$opacity
[1] 0



[[1]]$line
[[1]]$line$color
[1] "rgba(51,51,51,1)"

[[1]]$line$width
[1] 1.889764


[[1]]$name
[1] "Fair"

[[1]]$legendgroup
[1] "Fair"

[[1]]$showlegend
[1] TRUE

[[1]]$xaxis
[1] "x"

[[1]]$yaxis
[1] "y"

[[1]]$frame
[1] NA

[meldarionqeusse]

While I have not solved the issue above. I would like to add a different more complete solution to the provided by @isaaczhao23 in cases where there are boxplots with different colors.

df <- diamonds[sample(1:nrow(diamonds), size = 1000),]
#df <- subset(df,df$cut=="Fair")


p <- ggplot(df, aes(cut, price, color = cut)) + 
  geom_boxplot() 

# Need to modify the plotly object and make outlier points have opacity equal to 0
fig <- plotly_build(p)

fig$x$data <- lapply(fig$x$data, FUN = function(x){
  x$marker$outliercolor = x$line$color # When creating plot p with ggplot if you specify fill = cut use x$fill$color instead of $line$color
  x$marker$color = x$line$color # When creating plot p with ggplot if you specify fill = cut use x$fill$color instead $line$color
  x$marker$line = x$line$color # When creating plot p with ggplot if you specify fill = cut use x$fill$color instead $line$color
  return(x)
})
fig

Will produce this:

[lucazav]

I'm still having the issue of outliers shown after apply ggplotly() even if outlier.shape = NA is passed to geom_boxplot(). I'm using Plotly 4.9.4.1. Any chance to see this issue fixed? Thanks.

[lucazav]

In the meantime I solved the issue of hiding outliers using the following code:

library(purrr)

hideOutliers <- function(x) {  
  if (x$hoverinfo == 'y') {  
    x$marker = list(opacity = 0)
    x$hoverinfo = NA    
  }  
  return(x)  
}

p[["x"]][["data"]] <- map(p[["x"]][["data"]], ~ hideOutliers(.))

It works also with facets.

[swati-mahapatra]

The problem is that when you also have geom_jitter in the plot (in addition to geom_boxplot), the lapply part will remove all the points. Is there a way to selectively remove outliers that belong to geom_boxplot only?

You can use the code above and just index to the layer you want to remove, e.g. say the boxplot outliers are on the first layer.

p$x$data[1] <- lapply(p$x$data[1], FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

Thank you so much for this solution, @brshallo ! This fixed my problem after hours of looking for a fix!

[Tb0nes]

The problem is that when you also have geom_jitter in the plot (in addition to geom_boxplot), the lapply part will remove all the points. Is there a way to selectively remove outliers that belong to geom_boxplot only?

You can use the code above and just index to the layer you want to remove, e.g. say the boxplot outliers are on the first layer.

p$x$data[1] <- lapply(p$x$data[1], FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

As a novice, I'm having a hard time understanding what this code means. Would you or anyone else mind using this in a reproducible example? Thank you.

[LDSamson]

The problem is that when you also have geom_jitter in the plot (in addition to geom_boxplot), the lapply part will remove all the points. Is there a way to selectively remove outliers that belong to geom_boxplot only?

You can use the code above and just index to the layer you want to remove, e.g. say the boxplot outliers are on the first layer.

p$x$data[1] <- lapply(p$x$data[1], FUN = function(x){
  x$marker = list(opacity = 0)
  return(x)
})

As a novice, I'm having a hard time understanding what this code means. Would you or anyone else mind using this in a reproducible example? Thank you.

I used the information in this thread to create the following function, combining techniques described earlier for the specific use case that you mention. It might be helpful for others:

remove_boxplot_outliers <- function(fig){
  stopifnot("plotly" %in% class(fig))
  fig$x$data <- lapply(
    fig$x$data,
    \(i){
      if(i$type != "box") return(i)
      i$marker = list(opacity = 0)
      i$hoverinfo = "none"
      i
    }
  )
  fig
}

it only removes markers if type == "box". It also removes hoverinfo, otherwise you will still be able to see any hoverinfo of the outlier. Works also in faceted plots. Other layers, for example created with geom_jitter(), will be untouched. See below for an example

library(ggplot2)
library(plotly)
fig <- ggplotly({
  ggplot(iris, aes(factor(1), Petal.Width)) + geom_boxplot(outlier.shape = NA) + 
    geom_jitter(col = "red") + 
    facet_wrap(~Species)
})

Figure with outliers:

fig

Figure with outliers removed:

remove_boxplot_outliers(fig)