Adding Zoom to a Density Plot in Shiny - r

Question

I would like to be able to zoom in on the vlines in a density plot in shiny. I am using the iris dataset for reproducible purposes. The vlines are labeled with row names.
In my actual dataset, I have many vlines very close to each other. They are so close that I often cannot distinguish the row number labels. I would like to find a way to zoom in on the vline labels. I have attempted using a brush but that is not working.
library(tidyverse)
library(cluster)
library(shiny)
ui <- fluidPage({
pageWithSidebar(
headerPanel('Iris k-means clustering'),
sidebarPanel(
numericInput('clusters', 'Cluster count', 3, min = 1, max = 9)
),
mainPanel(
plotOutput('plot1',
dblclick = 'plot1_dblclick',
brush = brushOpts(
id = 'plot1_brush',
resetOnNew = TRUE
))
)
)
})
server <- function(input, output){
ClusterData <- reactive({
iris[,1:4]
})
# need to keep row numbers for outlier labels
ClusterData2 <- reactive({
ClusterData2 <- data.frame(ClusterData())
row.names(ClusterData2) <- 1:nrow(ClusterData2)
return(ClusterData2)
})
# scale the iris data
ScaledData <- reactive({
scale(ClusterData2())
})
# kmeans clustering
final <- reactive({
kmeans(ScaledData(), input$clusters, nstart = 25)
})
# find centers
states.centers <- reactive({
final()$centers[final()$cluster, ]
})
# find outliers
distances <- reactive({
sqrt(rowSums((ScaledData() - states.centers())^2))
})
# bind distances back to data
outliers <- reactive({
cbind(ClusterData(), Distance = distances())
})
# bind cluster number to data
clusterMember <- reactive({
cbind(outliers(), clusterNum = final()$cluster)
})
# turn into data frame
clusterMember2 <- reactive({
as.data.frame(clusterMember())
})
# find points that are their own cluster
dist0 <- reactive({
clusterMember() %>%
filter(distances() == 0)
})
# arrange distances largest to smallest
distArrange <- reactive({
clusterMember() %>%
arrange(desc(Distance))
})
# find top 5 outliers
filtTop5 <- reactive({
distArrange()[1:5,]
})
# bind outliers and single clusters together
AllOutliers <- reactive({
rbind(filtTop5(), dist0())
})
########## output plot
output$plot1 <- renderPlot({
ClusterData() %>%
ggplot(aes(x = Petal.Length)) +
geom_density(fill = "blue", alpha = 0.4) +
geom_vline(xintercept = AllOutliers()$Petal.Length) +
annotate("text", x = AllOutliers()$Petal.Length,
y = 0,
label = rownames(AllOutliers()),
hjust = 0.5,
vjust = -1)
})
######### zoom brush
observeEvent(input$plot1_dblclick, {
brush <- input$plot1_brush
if (!is.null(brush)) {
ranges$x <- c(brush$xmin, brush$xmax)
ranges$y <- c(brush$ymin, brush$ymax)
} else {
ranges$x <- NULL
ranges$y <- NULL
}
})
}
shinyApp(ui, server)
I am not invested in the above approach. Any suggestions would be welcome.
Thank you.

Answer 1

I advise you to try plotly package.
Modified code from your question:
library(tidyverse)
library(cluster)
library(shiny)
library(plotly)
ui <- fluidPage({
pageWithSidebar(
headerPanel('Iris k-means clustering'),
sidebarPanel(
numericInput('clusters', 'Cluster count', 3, min = 1, max = 9)
),
mainPanel(
plotlyOutput('plot1')
)
)
})
server <- function(input, output){
ClusterData <- reactive({
iris[,1:4]
})
# need to keep row numbers for outlier labels
ClusterData2 <- reactive({
ClusterData2 <- data.frame(ClusterData())
row.names(ClusterData2) <- 1:nrow(ClusterData2)
return(ClusterData2)
})
# scale the iris data
ScaledData <- reactive({
scale(ClusterData2())
})
# kmeans clustering
final <- reactive({
kmeans(ScaledData(), input$clusters, nstart = 25)
})
# find centers
states.centers <- reactive({
final()$centers[final()$cluster, ]
})
# find outliers
distances <- reactive({
sqrt(rowSums((ScaledData() - states.centers())^2))
})
# bind distances back to data
outliers <- reactive({
cbind(ClusterData(), Distance = distances())
})
# bind cluster number to data
clusterMember <- reactive({
cbind(outliers(), clusterNum = final()$cluster)
})
# turn into data frame
clusterMember2 <- reactive({
as.data.frame(clusterMember())
})
# find points that are their own cluster
dist0 <- reactive({
clusterMember() %>%
filter(distances() == 0)
})
# arrange distances largest to smallest
distArrange <- reactive({
clusterMember() %>%
arrange(desc(Distance))
})
# find top 5 outliers
filtTop5 <- reactive({
distArrange()[1:5,]
})
# bind outliers and single clusters together
AllOutliers <- reactive({
rbind(filtTop5(), dist0())
})
########## output plot
output$plot1 <- renderPlotly({
plot1 <- ClusterData() %>%
ggplot(aes(x = Petal.Length)) +
geom_density(fill = "#6495ed", alpha = 0.3) +
geom_vline(xintercept = AllOutliers()$Petal.Length, size = 0.5, colour = "#013220") +
annotate("text", x = AllOutliers()$Petal.Length,
y = 0,
label = rownames(AllOutliers()),
colour = "#d4240b",
size = 4)
ggplotly(plot1)
})
}
shinyApp(ui, server)
Output (zoomed with cluster count = 9):