Illustration of Randomisation Tests

server.R

library(shiny)

# Define server logic required to generate and plot a random distribution
shinyServer(function(input, output) {
  x1 <- c(3.36, 3.24, 5.16, 6.09, 3.08, 6.97, 4.85, 5.42, 5.15, 3.61, 4.95, 2.22)
  x2 <- c(2.72, 3.09, 1.49, 1.98, 0.85, 2.56, 2.14, 1.80, 0.57, 3.16, 0.90, 0.15)
  x <- c(x1, x2)
  mycol = c(rep("red", 12), rep("blue", 12))
  ind <- c(rep(1,12), rep(2,12))
  X <- matrix("numeric", 24, 1001)
  X[,1] <- c(1:24)
  for (j in 2:1001){
    X[,j] <- sample(c(1:24))
  }
  Y <- vector("numeric", 1001)
  for (j in 1:1001){
    Y[j] <- mean(x[as.numeric(X[1:12,j])]) - mean(x[as.numeric(X[13:24,j])])
  }
  hp <- hist(Y, plot = FALSE)
  #
  output$DataShuffle <- reactivePlot(function() {
    # this gives the numbers    
    r <- as.numeric(input$replicates)
    par(oma =c(0,0,0,0), xpd = NA)
    ##par(mfrow = c(2,1))
    layout(matrix(c(1,1,2,2), 2, 2, byrow = TRUE), widths=lcm(5), heights=lcm(10))
    #layout(matrix(c(1,1,2,3), 2, 2, byrow = TRUE))
    ##j <- 1
    plot(c(1,6), c(0,1), xaxt="n", yaxt="n", bty="n", type = "n", xlab = "", ylab = "", main = "")
    if (r ==1) {text(3,1.3,"Original Data", cex = 1.5)} else{text(3,1.3,paste("Shuffled Data:",(r-1)), cex = 1.5)}
    text(1:6, rep(1,6), x[as.numeric(X[c(1:6),r])], col = mycol[as.numeric(X[c(1:6),r])], cex = 1.5)
    text(1:6, rep(0.8,6), x[as.numeric(X[c(7:12),r])], col = mycol[as.numeric(X[c(7:12),r])], cex = 1.5)
    text(1:6, rep(0,6), x[as.numeric(X[c(13:18),r])], col = mycol[as.numeric(X[c(13:18),r])], cex = 1.5)
    text(1:6, rep(0.2,6), x[as.numeric(X[c(19:24),r])], col = mycol[as.numeric(X[c(19:24),r])], cex = 1.5)
    t1 <- bquote(bar(x)[1] - bar(x)[2] ==.(round(mean(x[as.numeric(X[c(1:12),r])]) - mean(x[as.numeric(X[c(13:24),r])]),2)) )
    text(3.5, 0.5, t1, cex=2)
  })
  
  output$NullHypHist <- reactivePlot(function(){
    r <- as.numeric(input$replicates)
    par(oma =c(0,0,0,0), las = 1)
    ##par(mfrow = c(2,1))
    hist(Y[1:r], xlim = c(-5, 5), breaks = hp$breaks, ylim = c(0, max(hp$counts)), main = "Null Hypothesis Distribution")
    points(Y[r], 0, pch = 16, col = "red")
    #acplot(1,1)
  })
})

ui.R

library(shiny)

# Define UI for application that plots random distributions 
shinyUI(pageWithSidebar(
  
  # Application title
  headerPanel("Simple randomisation test"),
  
  # Sidebar with a slider input for number of observations
  sidebarPanel(
    sliderInput("replicates", 
                "Number of shuffles:", 
                min = 1, 
                max = 1000, 
                value = 1)
  ),
  
  # Show a plot of the generated distribution
  mainPanel(
    #conditionalPanel(condition = "input.DataShuffle",
    plotOutput(outputId = "DataShuffle"),
    
    #conditionalPanel(condition = "input.NullHypHist",
    plotOutput(outputId = "NullHypHist")
    
  )
))

This is version 0.1 of an illustration of Randomisation tests.

Things to improve

1. Upload user data (maybe with some row limit)
2. Better illustration of critical region / observed test statistic (I think that's missing altogether at the moment)

Actually, the slider bar doesn't work too well (used to work fine as a tkPanel app) and you don't get to see it moving in steps of 1. Maybe this is better as an animation in any case and you set the delay time/?????