In dtkaplan/mdsint: Internal resources for Modeling for Data Science

library(learnr)
library(shiny)
library(mdsint)
library(ggformula)
library(mosaic)
library(mosaicModel)
knitr::opts_chunk$set(echo = FALSE)

checkboxInput("bygroups", "Show drivers vs conductors", value = FALSE)
checkboxInput("logistic", "Use logistic regression", value = FALSE)
plotOutput("riskplot", brush = brushOpts(id = "age", fill = "#ccc", direction = "x"))

data(Bus_workers, package = "mdsint")
get_formula <- reactive({
    if (input$bygroups) { event == "attack" ~ age + job }
    else { event == "attack" ~ age }
})
# handle initiation and make sure the age range is large enough
min_age <- reactive({if (is.null(input$age)) 45 else pmin(60, pmax(35, input$age$xmin))})
max_age <- reactive({if (is.null(input$age)) 55 else pmin(65, pmax(min_age()+5, input$age$xmax))})

get_data <- reactive({
  Bus_workers[(Bus_workers$age >= min_age()) & 
                  (Bus_workers$age <= max_age()), ]
})
get_model <- reactive({
  f <- get_formula()
  This_data <- get_data()

  if (input$logistic)
    glm(f, data = This_data, family = binomial)
  else
    lm(f, data = This_data)
})
get_baseline_curve <- reactive({
  f <- get_formula()

  mod <- 
    if (input$logistic)
      glm(f, data = Bus_workers, family = binomial)
    else
      lm(f, data = Bus_workers)
  mod_eval(mod, age = 35:64)
})
output$riskplot <- renderPlot({
  mod <- get_model()
  curve <- get_baseline_curve()
  This_data <- get_data()
  mod_plot(mod, interval = "confidence", data = This_data) %>% 
    gf_line(data = curve, model_output ~ age, alpha = 0.5)  %>%
    gf_lims(x = c(35,65), y = c(-0.005, .025)) %>%
    gf_labs(y = "Prob. of heart attack in the next year")
})