Add a regression line to my scatterplot in r shiny

Question

So I have built a shiny app that allows you to create a scatterplot based on any combination of my variables from my dataset. It also outputs a summary of a regression model on a separate tab.

Basically I want an option to add a regression line to my scatterplot. This would be based on the inputs chosen from the UI to create the scatterplot. I have made a few attempts but nothing successful. Here is the working code I have so far:

  pageWithSidebar(
    titlePanel("Plotting weather trends 2010 - 2014"),
    sidebarPanel(
      selectInput("xCol", "Please select an x variable", names(DF2)),
      selectInput("yCol", "Please select a y variable", names(DF2)),
      checkboxInput("line", "Show regression line?", value = TRUE),
      selectInput("plot_type", "Select a secodary plot for X variable", choices = c("Boxplot", "Histogram"))),
    mainPanel(tabsetPanel(type = "tabs",
                          tabPanel("Plots", (plotOutput("plot1")), plotOutput("plot2")),
                          tabPanel("Regression Model Summary", verbatimTextOutput(outputId = "RegSum"))
    )
    )
  ))


server = function(input, output){ 

  DF3 = reactive({DF2[, c(input$xCol, input$yCol)]})



  output$plot1 = renderPlot({plot(DF3(),
                                  main = "Histogram of select variables")})



  output$plot2 = renderPlot({ 
    if (input$plot_type == "Boxplot") {boxplot(DF2[,input$xCol], main = "boxplot of X variable")}
    if (input$plot_type == "Histogram") {hist(as.numeric(unlist(DF2[,input$xCol])),main = "Histogram of X variable", xlab = "X variable")}
  })

  lm1 <- reactive({lm(reformulate(input$xCol, input$yCol), data = DF2)})
  output$RegSum <- renderPrint({summary(lm1())})

}

shinyApp(ui = ui, server = server)

Previous attempts have included adding an abline(lm) function intaking the inputs from the UI but to no avail.

Any help would be appreciated.

Answer 1

You were on the right way with abline . I guess your problem was how you specified the lm function: If a user selects two variables "xCol" and "yCol", these are passed in quotes to shiny's server. However, the lm function expects a formula notation of the form y ~ x . To get around this problem I would write lm(get(input$yCol) ~ get(input$xCol), data=DF3()) . This way R searches within the data set rather than the global environment.

Here is a reproducible example base on the built-in mtcars dataset:

library(shiny)

ui <- fluidPage(
  selectInput("xCol", "Please select an x variable", names(mtcars)),
  selectInput("yCol", "Please select a y variable", names(mtcars)),
  checkboxInput("line", "Show regression line?", value = TRUE),
  plotOutput("plot")
)

server <- function(input, output, session) {

  data <- reactive({mtcars[, c(input$xCol, input$yCol)]})

  output$plot <- renderPlot({
    plot(data())
    if(input$line) {
      model <- lm(get(input$yCol) ~ get(input$xCol), data=data())
      abline(model)
    }
  })
}

shinyApp(ui, server)