ui.R

library(shiny)
library(markdown)
library(shinydashboard)
library(DynTxRegime)
library(DT)
library(rgenoud)

source("./modules/uploadDat.R")

header <-
  dashboardHeader(title = "Dynamic Treatment Regimes",titleWidth = 350, disable = FALSE)

sidebar <- dashboardSidebar(
  width = 350,
  tags$head(
    tags$link(rel = "stylesheet", type = "text/css", href = "custom.css"),
    tags$script(src = "custom.js")
  ),
  sidebarMenu(
    menuItem(
      text = "Causal Inference Background",
      menuItem(text = "Motivation", tabName = "mot"),
      menuItem(
        text = "Causal Effect",
        menuSubItem(text = "Individual Causal Effect", tabName = "ice"),
        menuSubItem(text = "Average Causal Effect", tabName = "ace"),
        menuSubItem(text = "Summary", tabName = "sce")
      ),
      menuItem(
        text = "Point Exposure Studies",
        menuSubItem(text = "Randomized Studies", tabName = "rs"),
        menuSubItem(text = "Observational studies", tabName = "os"),
        menuSubItem(text = "Notation", tabName = "not")
      ),
      menuItem(
        text = "Three Necessary Assumptions",
        menuSubItem(text = "Necessary Assumptions", tabName = "na1"),
        menuSubItem(text = "Assumption 1 : SUTVA", tabName = "sutva"),
        menuSubItem(text = "Assumption 2 : NUC", tabName = "nuc"),
        menuSubItem(text = "Assumption 3 : Positivity", tabName = "pos"),
        menuSubItem(text = "Necessary Assumptions (Cont'd)", tabName = "na2")
      ),
      menuItem(text = "Optimal Treatment Regimes", tabName = "otr")
    ),
    
    menuItem(text = "Dataset", tabName = "dataone"),
    
    menuItem(
      text = "Single Stage Methods",
      menuItem(
        text = "Outcome Regression",
        menuSubItem(text = "Outcome Regression", tabName = "or"),
        menuSubItem(text = "Case Study", tabName = "orc")
      ),
      menuItem(
        text = "Augmented Inverse Probability Weighted Estimator",
        menuSubItem(text = "AIPWE", tabName = "aipwe"),
        menuSubItem(text = "Case study", tabName = "ac")
      ),
      menuItem(
        text = "Classification Method",
        menuSubItem(text = "Classification Method", tabName = "class"),
        menuSubItem(text = "Case Study", tabName = "cc")
      ),
      menuItem(text = "Hands-on", tabName = "handson")
    ),
    menuItem(text = "Reference", tabName = "ref")
  )
)

body <- dashboardBody(
  tabItems(
    tabItem(tabName = "mot",
            withMathJax(
              includeMarkdown("./www/background/begin.Rmd")
              )),
    tabItem(tabName = "ice",
            withMathJax(
              includeMarkdown("./www/background/effect_1.Rmd")
              )),
    tabItem(tabName = "ace",
            withMathJax(
              includeMarkdown("./www/background/effect_2.Rmd")
            )),
    tabItem(tabName = "sce",
            withMathJax(includeMarkdown("./www/background/effect_3.Rmd")
            )),
    tabItem(tabName = "rs",
            withMathJax(includeMarkdown("./www/background/point_1.Rmd")
            )),
    tabItem(tabName = "os",
            withMathJax(includeMarkdown("./www/background/point_2.Rmd")
            )),
    tabItem(tabName = "not",
            withMathJax(includeMarkdown("./www/background/notation.Rmd")
            )),
    tabItem(tabName = "na1",
            withMathJax(includeMarkdown("./www/background/assumption1.Rmd")
            )),
    tabItem(tabName = "sutva",
            withMathJax(
              includeMarkdown("./www/background/sutva.Rmd")
            )),
    tabItem(tabName = 'nuc',
            withMathJax(
              includeMarkdown("./www/background/nuc.Rmd")
            )),
    tabItem(tabName = 'pos',
            withMathJax(
              includeMarkdown("./www/background/positivity.Rmd")
            )),
    tabItem(tabName = "na2",
            withMathJax(
              includeMarkdown("./www/background/assumption2.Rmd")
            )),
    tabItem(tabName = "otr",
            withMathJax(
              includeMarkdown("./www/background/optimal.Rmd")
            )),
    tabItem(tabName = "dataone",
            uiOutput("dataone")),
    tabItem(tabName = "or",
            withMathJax(includeMarkdown("./www/outcome_regress.Rmd"))),
    tabItem(tabName = "orc",
            uiOutput("orc")),
    tabItem(tabName = "aipwe",
            withMathJax(includeMarkdown("./www/aipwe.Rmd"))),
    tabItem(tabName = "ac",
            uiOutput("ac")),
    tabItem(tabName = "class",
            withMathJax(includeMarkdown("./www/class.Rmd"))),
    tabItem(tabName = "cc",
            uiOutput("cc")),
    tabItem(tabName = "ref",
            withMathJax(includeMarkdown("./www/background/reference.Rmd"))),
    
    tabItem(
      tabName = "handson",
      tabsetPanel(
        type = "tabs",
        tabPanel("Upload Data",
          sidebarLayout(
            sidebarPanel(
              h3("Upload Your Data:"),
                p("You may load a small sized csv/text file (less
                  that 5Mb) to interactiely learn about the outcome regression
                  method."),
                  csvFileInput("datafile", "User data (.csv format)")),
                  mainPanel(dataTableOutput("table"))
            )
          ),
        tabPanel(
          "Data Explore",
          fluidPage(
            h4("Data Explore with Scatter Plots"),
            p("You may use scatter plots to visualize your uploaded dataset.")
          ),
          actionButton("getplot", "Go"),
          plotOutput("plot0")
        ),
        tabPanel("Methods",
            fluidPage(
              fluidRow(wellPanel(  # radio buttons for conditionalPanel, allow user to choose methods
              radioButtons("radio", label = h4("Choose a method"),
                           choices = list("Outcome Regression" = 1, "AIPWE" = 2, 
                                          "Classification Method" = 3), 
                           selected = 1, inline=TRUE))),
              # Only show this panel if the plot type is a histogram
              conditionalPanel(
                condition = "input.radio == '1'",
                fluidPage(
                  sidebarLayout(
                    sidebarPanel(# let user choose variable and models
                      uiOutput("varTrtR"),
                      uiOutput("varResponseR"),
                      uiOutput("varMainR"),
                      uiOutput("solverMainR"),
                      uiOutput("varContR"),
                      uiOutput("solverContR"),
                      width =4
                    ),
                    mainPanel(
                      fluidPage(
                        h2("Outcome Regression"),
                        h4("Build your model objects for main effect and contrast and fit"),
                        p("To build model objects for main effect (moMain) and
                          contrast (moCont). You may specify your treatment variable and response
                          varible below. You may also choose the covariates to include in and
                          solver for each model. Here, we choose the solver to be 'lm'. Yet,
                          other methods, such as glm, can be used for each model independently.
                          This step can be achieved using the function buildModelObj(), followed by
                          call the qLearn() function to fit the models together."),
                        h4("Fit the model specified"),
                        actionButton("getmodelR", "GO"),
                        # h4("The corresponding code:"),
                        # textOutput("myMainR"),
                        # textOutput("myContR"),
                        # textOutput("myFitQ1"),
                        h4("Extract the estimated coefficients"),
                        p("We may extract the estimated coefficients by  function coeff()"),
                        #code("coef(fitQ1)"),
                        tableOutput("coeffTbl"),
                        h4("Diagnostic plots"),
                        p("We may use the function plot() to obtain the diagnostic plots for fitted model"),
                       # code("plot(fitQ1)"),
                        fluidRow(
                          column(6, plotOutput("plot1")),
                          column(6, plotOutput("plot2"))
                          ),
                        fluidRow(
                          column(6, plotOutput("plot3")),
                          column(6, plotOutput("plot4"))
                          )
                        )
                      )
                    )
                  )
                ),
              conditionalPanel(
                condition = "input.radio == '2'",
                fluidPage(
                   #interaction not included yet
                  sidebarLayout(
                    sidebarPanel(
                      uiOutput("varTrtA"),
                      uiOutput("varResponseA"),
                      uiOutput("varPropA"),
                      # uiOutput("solverProp"),
                      uiOutput("varMainA"),
                      #  uiOutput("solverMain"),
                      uiOutput("varContA"),
                      #  uiOutput("solverCont"),
                      uiOutput("varRuleA")
                      #uiOutput("solverCont")
                    ),
                    mainPanel(
                      fluidPage(
                        h2("Augmented Inverse Probility Weighted Estimator"),
                        h4("Specify treatment variable and build your modeling objects."),
                        p("Before running the optimalSeq function, we need to for the 
                          treatment variable. This tells optimalSeq 'optimalSeq' which 
                          columns of data correspond to treatments, and build the modeling 
                          objects for propensity score (moPropen) and conditional expectations 
                          of the main and contrast effect (expec.main and expec.cont).
                          You may specify your treatment variable and response varible below. 
                          You may also choose the covariates to include in and solver for each model. 
                          Here, we choose the solver to be glm for modeling propensity score and 
                          gl for conditional expectations."),
                        h4("Estimate the optimal regime"),
                        actionButton("getmodelA", "GO"),
                        uiOutput("aipwe")
                        # uiOutput("ipwe")
                        # p("The corresponding code for build models:"),
                        # textOutput("myPropA"),
                        # textOutput("myMainA"),
                        # textOutput("myContA")
                      )
                     )
                   )
                  )
                ),
              # Only show this panel if Custom is selected
              conditionalPanel(
                condition = "input.radio == '3'",
                
                fluidPage(

                  
                  sidebarLayout(
                    sidebarPanel(
                      uiOutput("varTrtC"),
                      uiOutput("varResponseC"),
                      uiOutput("varPropC"),
                      uiOutput("varMainC"),
                      uiOutput("varContC"),
                      uiOutput("varClass")
                    ),
                    mainPanel(
                      fluidPage(
                        h2("Classification"),
                        h4("Specify treatment variable and build your modeling objects."),
                        p("Before running the optimalClass function, we need to for the 
                          treatment variable. This tells optimalClass 'optimalClass' which 
                          columns of data correspond to treatments, and build the modeling 
                          objects for propensity score (moPropen) and conditional expectations
                          (moMain and moCont), You may specify your treatment variable 
                          and response varible You may also choose the covariates to include
                          in and solver for each model. You may also choose variables to specify
                          the class of regimes. The class of regime is restricted to be linear 
                          decision rule"),
                        h6("Estimate the optimal regime"),
                        actionButton("getmodelC", "GO"),
                        uiOutput("optClass")
                        )
                      )
                    )
                  )
                
          )
        )
      )
    )
  )
))

    








dashboardPage(header, sidebar, body, tags$head(tags$style(
  HTML(
    '
    .skin-blue .main-sidebar {
    background-color: #666666;
    }
    .skin-blue .sidebar-menu>li.active>a, .skin-blue .sidebar-menu>li:hover>a {
    background-color: #444444;
    }
    '
  )
  )))