Chi-squared Calculator
Chi-squared Probabilities.
In this tool, you input a \(\chi^2\) score and the degrees of freedom, and get a corresponding \(p\)-value.
#| standalone: true
#| viewerHeight: 500
library(shiny)
ui <- fluidPage(
inputPanel(
selectInput("tails", label = "Which tails",
choices = c("Upper tail: Pr(x^2 < X^2)"="upper",
"Lower tail: Pr(X^2 < x^2)"="lower"),
selected = "upper"),
numericInput("x2", label = "chi-squared value:", value=2),
numericInput("df", label = "Degrees of Freedom", value =1)
),
mainPanel(
plotOutput("probplot")))
server <- function (input,output) {
output$probplot <- renderPlot({
q <- input$x2
df <- input$df
p <- switch(input$tails,
upper=1-pchisq(q,df),
lower=pchisq(q,df))
xl <- round(qchisq(.999,df),1)
curve(dchisq(x,df),main=paste("Probability of shaded region = ",round(p,3)),
sub=paste("chi-squared = ",round(q,3)),
xlim = c(0,xl),yaxt="n",cex=3,cex.lab=2,cex.main=2,ylab="",xlab="Chi-squared")
switch(input$tails,
upper={
cord.xu <- c(q,seq(q,xl,0.01),xl)
cord.yu <- c(0,dchisq(seq(q,xl,0.01),df),0)
polygon(cord.xu,cord.yu,col='plum')
axis(1,q,paste(round(q,3)),cex.axis=2)
},
lower={
cord.xl <- c(0,seq(0,q,0.01),q)
cord.yl <- c(0,dchisq(seq(0,q,0.01),df),0)
if(!is.finite(cord.yl[2])) cord.yl[2] <- cord.yl[3]
polygon(cord.xl,cord.yl,col='plum')
axis(1,q,paste(round(q,3)),cex.axis=2)
})
})
}
shinyApp(ui=ui,server=server)
Chi-square Quantiles (Critical values).
In this tool, you input a probability and degrees of freedom, and get a corresponding \(X^2\) score.
#| standalone: true
#| viewerHeight: 500
library(shiny)
ui1 <- fluidPage(
inputPanel(
selectInput("tails1", label = "Which tails",
choices = c("Upper tail: Pr(x^2 < X^2)"="upper",
"Lower tail: Pr(X^2 < x^2)"="lower"),
selected = "upper"),
numericInput("pp", label = "Probability of shaded region:", value=0.05, min=0, max=1),
numericInput("df1", label = "Degrees of Freedom", value =1)
),
mainPanel(
plotOutput("quantplot")))
server1 <- function (input,output) {
output$quantplot <- renderPlot({
pp <- as.numeric(input$pp)
df1 <- as.numeric(input$df1)
tails1 <- input$tails1
if (is.null(tails1)) tails1 <- "upper"
qq <- switch(tails1,
upper=qchisq(1-pp,df1),
lower=qchisq(pp,df1))
xl1 <- round(qchisq(.999,df1),1)
curve(dchisq(x,df1),main=paste("Probability of shaded region = ",round(pp,3)),
sub=paste("X2(",df1,") = ",round(qq,3)),
xlim = c(0,xl1),yaxt="n",cex=3,cex.lab=2,cex.main=2,ylab="",xlab="Chi-squared")
switch(tails1,
upper={
cord.xu1 <- c(qq,seq(qq,xl1,0.01),xl1)
cord.yu1 <- c(0,dchisq(seq(qq,xl1,0.01),df1),0)
polygon(cord.xu1,cord.yu1,col='plum')
axis(1,qq,paste(round(qq,3)),cex.axis=2)
},
lower={
cord.xl1 <- c(0,seq(0,qq,0.01),q)
cord.yl1 <- c(0,dchisq(seq(0,qq,0.01),df1),0)
if(!is.finite(cord.yl[2])) cord.yl[2] <- cord.yl[3]
polygon(cord.xl1,cord.yl1,col='plum')
axis(1,qq,paste(round(qq,3)),cex.axis=2)
})
})
}
shinyApp(ui=ui1,server=server1)
For \(\chi^2\) tests, one almost always looks at upper tail, and rarely look at the lower tail (does the data fit better than expected). Two-tailed tests are never (“What never?”, “No Never!”, “What Never?”, “Well hardly ever!”, Gilbert & Sulivan, HMS Pinfore) done.