如何在R中定制Plotly箱线图的悬停文本

Question

我知道如何在plotly中自定义散点图的悬停文本，但箱线图不接受“text”属性。出现警告信息：“'box'对象没有这些属性：'text'”。我有超过300个x轴变量，并且有两组(A或B)中的编号样本(1-50)，我想将它们一起绘制在同一个箱线图中，然后当移动光标到异常值上时，我想通过悬停文本区分样本编号和组别。我想要自定义数据标签而不是自动四分位标签。在plotly箱线图中是否可能实现？

library(plotly) 
library(magrittr)

plot_ly(melt.s.data, 
          x = ~variable, 
          y = ~value,
          type = 'box', 
          text = ~paste("Sample number: ", Sample_number, 
                       '<br>Group:', Group)) %>% 
        layout(title = "Individual distributions at each x")

这里有一些样本数据，仅显示了5个x变量（但代码应该可以扩展到我的300个）...

#sample data
set.seed(456)
#Group A
sample.data_a <- data.frame(Class = "red", Group = "A",
                            Sample_number = seq(1,50,by=1), 
                            x1= rnorm(50,mean=0, sd=.5), 
                            x2= rnorm(50,mean=0.5, sd=1.5), 
                            x3= rnorm(50,mean=5, sd=.1), 
                            x4= rnorm(50,mean=0, sd=3.5),
                            x5= rnorm(50,mean=-6, sd=.005))
#Group B
sample.data_b <- data.frame(Class = "red", Group = "B",
                            Sample_number = seq(1,50,by=1), 
                            x1= rnorm(50,mean=0, sd=5.5), 
                            x2= rnorm(50,mean=0.5, sd=7.5), 
                            x3= rnorm(50,mean=5, sd=.01), 
                            x4= rnorm(50,mean=0, sd=.5),
                            x5= rnorm(50,mean=-6, sd=2.05))

#row Bind groups 
sample.data <- rbind(sample.data_a, sample.data_b)

#melting data to have a more graphable format
library(reshape2)
melt.s.data<-melt(sample.data, id.vars=c("Class", "Group","Sample_number"))

以下是类似的问题：

• 这里 看起来似乎不可能实现。
• 这个问题类似，但只想添加相关的四分位信息。
• 而这个问题仅涉及在plotly盒形图中单个数据点的文本标注。

- andemexoax

---

请提供一个可重现的示例。 - MLavoie

@MLavoie 我添加了可重现示例的数据。 - andemexoax

andemexoax，你很友善地接受了我的答案，谢谢。请查看我的博客，以获取另一个更好的解决方案（因为它确保工具提示永远不会超出图表范围）。 - Stéphane Laurent

Answer 1

可以使用Shiny实现。

library(plotly)
library(shiny)
library(htmlwidgets)

# Prepare data ----
set.seed(456)
#Group A
sample.data_a <- data.frame(Class = "red", Group = "A",
                            Sample_number = seq(1,50,by=1), 
                            x1= rnorm(50,mean=0, sd=.5), 
                            x2= rnorm(50,mean=0.5, sd=1.5), 
                            x3= rnorm(50,mean=5, sd=.1), 
                            x4= rnorm(50,mean=0, sd=3.5),
                            x5= rnorm(50,mean=-6, sd=.005))
#Group B
sample.data_b <- data.frame(Class = "red", Group = "B",
                            Sample_number = seq(1,50,by=1), 
                            x1= rnorm(50,mean=0, sd=5.5), 
                            x2= rnorm(50,mean=0.5, sd=7.5), 
                            x3= rnorm(50,mean=5, sd=.01), 
                            x4= rnorm(50,mean=0, sd=.5),
                            x5= rnorm(50,mean=-6, sd=2.05))
#row Bind groups 
sample.data <- rbind(sample.data_a, sample.data_b)
#melting data to have a more graphable format
melt.s.data <- reshape2::melt(sample.data, 
                              id.vars=c("Class", "Group", "Sample_number"))

# Plotly on hover event ----
addHoverBehavior <- c(
  "function(el, x){",
  "  el.on('plotly_hover', function(data) {",
  "    if(data.points.length==1){",
  "      $('.hovertext').hide();",
  "      Shiny.setInputValue('hovering', true);",
  "      var d = data.points[0];",
  "      Shiny.setInputValue('left_px', d.xaxis.d2p(d.x) + d.xaxis._offset);",
  "      Shiny.setInputValue('top_px', d.yaxis.l2p(d.y) + d.yaxis._offset);",
  "      Shiny.setInputValue('dy', d.y);",
  "      Shiny.setInputValue('dtext', d.text);",
  "    }",
  "  });",
  "  el.on('plotly_unhover', function(data) {",
  "    Shiny.setInputValue('hovering', false);",
  "  });",
  "}")

# Shiny app ----
ui <- fluidPage(
  tags$head(
    # style for the tooltip with an arrow (http://www.cssarrowplease.com/)
    tags$style("
               .arrow_box {
                    position: absolute;
                  pointer-events: none;
                  z-index: 100;
                  white-space: nowrap;
                  background: CornflowerBlue;
                  color: white;
                  font-size: 13px;
                  border: 1px solid;
                  border-color: CornflowerBlue;
                  border-radius: 1px;
               }
               .arrow_box:after, .arrow_box:before {
                  right: 100%;
                  top: 50%;
                  border: solid transparent;
                  content: ' ';
                  height: 0;
                  width: 0;
                  position: absolute;
                  pointer-events: none;
               }
               .arrow_box:after {
                  border-color: rgba(136,183,213,0);
                  border-right-color: CornflowerBlue;
                  border-width: 4px;
                  margin-top: -4px;
               }
               .arrow_box:before {
                  border-color: rgba(194,225,245,0);
                  border-right-color: CornflowerBlue;
                  border-width: 10px;
                  margin-top: -10px;
               }")
  ),
  div(
    style = "position:relative",
    plotlyOutput("myplot"),
    uiOutput("hover_info")
  )
)

server <- function(input, output){
  output$myplot <- renderPlotly({
    plot_ly(melt.s.data, 
            type = "box", 
            x = ~variable, y = ~value, 
            text = paste0("<b> group: </b>", melt.s.data$Group, "<br/>",
                          "<b> sample: </b>", melt.s.data$Sample_number, "<br/>"),
            hoverinfo = "y") %>%
      onRender(addHoverBehavior)
  })
  output$hover_info <- renderUI({
    if(isTRUE(input[["hovering"]])){
      style <- paste0("left: ", input[["left_px"]] + 4 + 5, "px;", # 4 = border-width after
                      "top: ", input[["top_px"]] - 24 - 2 - 1, "px;") # 24 = line-height/2 * number of lines; 2 = padding; 1 = border thickness
      div(
        class = "arrow_box", style = style,
        p(HTML(input$dtext, 
               "<b> value: </b>", formatC(input$dy)), 
          style="margin: 0; padding: 2px; line-height: 16px;")
      )
    }
  })
}

shinyApp(ui = ui, server = server)

- Stéphane Laurent

Answer 2

一个可能的解决方案是使用ggplot2包，并向你的箱线图添加一个不可见的散点图：

library(ggplot2)
library(plotly)

gg_box <- melt.s.data %>%
  ggplot(aes(x=variable, y=value, text=paste("Group:",Group, "\n",
                                            "Class:", Class))) +
  geom_boxplot()+

  #invisible layer of points
  geom_point(alpha = 0)

gg_box %>%
  ggplotly() 

您需要稍微移动一下光标，才能看到额外的标签。

- Hamideh