readme_shiny.md

Tutorial on how to reproduce and modify the Shiny App for the Deworming Open Policy Analysis

The shiny app for the deworming OPA contains three files: ui.R, server.R, and all_analysis.R.

• 01_final_opa file contains the full walkthrough for the analysis, which is then represented in a more code-friendly way in all_analysis.R
• The data is all_analysis.R is passed to...
  • ui.R, which is the user interface
  • server.R, which manages the server components of the shiny app

The purpose of the shiny app is to allow the user to experiment with the data to see the changes that those differences make, in order to better understand the analysis. It turns policy analysis, which is often complex and hard to understand, into an easily readable and understandable format.

Your setup to reproduce and modify the ui.R of the deworming OPA

• To run the shiny app (SA) you will need to install R and RStudio. The relevant file can be found by running ui.R.

• Click 'Run App' located at the top of RStudio. This will open the SA.

• The SA has three tabs to explore

  • Main Policy Estimate, which describes the results. There is no interactive capabilities on this tab.
  • Key Assumptions, where you can adjust 'policy estimate', 'unit costs in new country', and 'prevelance in the new region'.
  • Main Assumptions, where you can adjust 'number of simulations' and 'policy estimate', as well as choose between 'Data' and 'Research'
    • Data
      • Gov Bonds (and SD)
      • Inflation (and SD)
      • Exchange rate (and SD)
      • Costs of treatment (and SD)
      • Years of treatment in original study (and SD)
      • Years of treatment in new setting (and SD)
      • Costs adjustments (and SD)
    • Research
      • Take-up (and SD)
      • Take-up with no subsidy (and SD)
      • Lambda 1_1_new (and SD)
      • Lambda 1_2_new (and SD)
      • Lambda 1_3_new (and SD)
      • Prevalence in original study (and SD)

Beginning

• The code begins with a list of all relevant libraries
• Then, it states the unit costs and the prevalences of 4 countries: India, Kenya, Nigeria, and Vietnam
• At the top of the SA, it says "Open Policy Analysis for Deworming Interventions: Open Output Component"
  • This is created using navbarPage

The tabs

• All three tabs are created using tabPanel.
  • Then, use sidebarPanel, which creates the panel on the left side of the screen. For tab titled "Main Policy Estimate", this only consists of text and links.
    • fluidrow creates rows, such as "Unit costs in new country" in "Key Assumptions".
  • mainPanel creates the graph seen on the right

The functions

• shinyUI -> creates a user interface
• fluidPage -> creates a page
• navbarPage -> creates a page with a navigation bar at the top
• tabPanel -> creates a tab panel, takes in a title (i.e. "Main Policy Estimate") and a value
• sidebarPanel -> creates a sidebar that has input controls
• tags$a -> creates a link
• mainPanel -> creates a main panel of output elements
• plotOutput -> creates a plot
• checkboxInput -> creates a checkbox used to specify logical values
• numericInput -> creates an input control for numeric values
• withMathJax -> loads in MathJax, a JavaScript library
• useShinyjs -> loads in Shinyjs, an r package
• helpText -> creates text to help explain some element of the SA
• selectInput -> creates a list of choices
• sliderInput -> creates a slider widget containing numbers in a range
• tabsetPanel -> creates a tabset containing elements from tabPanel
• uiOutput -> tells Shiny where the controls should be rendered

Your setup to reproduce and modify the server.R file of the deworming OPA

The functions

• sim.data -> returns a random sample from some distributions
• observeEvent -> chooses which slider inputs to show and which sliders not to show
• ggplot -> data visualization package
  • geom_density -> computes and draws density estimate
  • geom_vline -> annotates the plot with vertical lines
  • xlim -> sets the min and max boundaries of the plot in the x-direction
  • guides -> the guide is set scale-by-scale for each scale
  • annotate -> adds small annotations, i.e. text labels
  • theme -> sets the theme of the plot by costumizing all non-data components

Walkthrough

• reactive.data1 is an R expression that uses a widget input and outputs a value, and in this case the input is sim_data.
• sim_data sets variable names by turning the parameters used in ui.R to numeric factors
• The following section of server.R is if statements that are used to show which parameters to hide for different inputs of 'policy estimate'.
• Then, the code adds helps text to the SA depending on the input for 'policy estimate'.
• The next section defines a function that generates policy estimate plots
  • If statements specify what texts to display and what selectInput to call for each of the three different plots (Main, ka, all)
  • Then, the code constructs the plot for all three tabs
    • All have the same axes
    • All are density plots
    • All share the same plot structure

How to add a new slider to the UI

• In order to add a slider, you will want to use the sliderInput function in ui. R. sliderInput is a widget that creates a slider across a range. It has five inputs: inputId, label, min, max, and value. Below is an example:

  sliderInput("param3", label = "Inflation (\\( \\pi \\) ) = ", min = 0.001, max = 0.2, value = inflation_so)

• Here, param3 is the inputID, Inflation (π) = is the label, min is 0.001, max is 0.2, and the value is inflation_so

• To add a new slider, you will first need to figure out what the inputID should be.

  • First, you will need to add the inputID to the server file. Define a variable in the function sim.data1 in server.R. For the example above this would be inflation_var2 = as.numeric(input$param3).
  • simdata1 is contained in the reactive.data1.

• Then, if you don't want the slider to appear for certain inputs of 'policy estimate', insert it in else if(input$policy_est == "What policy estimate you don't want the slider to appear in").

• It goes in the list_hide function within the else if. Once that has been added, that variable, in this case inflation_var2, should be added to simdata1 in all_analysis.R. This deals directly with the original analysis.

• EXPLAIN: knitr::purl("code/01_final_opa.Rmd", "code/shiny_app/all_analysis.R")

• If the slider you want to add is already in all_analysis.R, make sure to use the name defined in all_analysis.R, then add it to server.R, and then ui.R.

• To change the value, which is inflation.so in the above example, modify it in the "Data" section of all_analysis.R.

The chart below shows the way that the different components of the OPA interact with each other

The line, knitr::purl("code/01_final_opa.Rmd", "code/shiny_app/all_analysis.R") converts the R markdown in 01_final_opa.Rmd to R script in all_analysis.R. This means that when data is updated or added in 01_final_opa.Rmd, it will update the all_analysis.R file.

Your setup to deploy and publish the Shiny App for the Deworming Open Policy Analysis

One effective way to publicly share and show your shiny app (SA) for the deworming OPA is using shinyapps.io, an online hosting service RStudio provides. You can deploy and publish your SA by following the steps below.

1. Create a shinyapps.io account and connect it with your RStudio.
2. Make sure your codes can work in the shinyapps.io cloud.

1. Create a shinyapps.io account and connect it with your RStudio

*Please skip this step if you already have a shinyapps.io account.

• In ui.R, click the icon to the right to "Run App" and select "Manage Accounts".

• In "Publishing" option, click "Connect..." to the right to "Publishing Accounts".

• Select "ShinyApps.io" and click the "Get started here" link.

• After signing up with shinyapps.io, following the instruction of RStudio, get, copy and paste the token. Finally, click "Connect Account".

  

2. Make sure your codes can work in the shinyapps.io cloud

• Put the folders containing the related files in your local environment.

• In Line 9 in all_analysis.R, turn "TRUE" into "FALSE" in the if function.

  

• In Lines 55 and 56 in all_analysis.R, inactivate functions by putting "#" at the beginning.

  

• Then, in ui.R, click the icon to the right to "Run App" and select "Other Destination...". After making sure that all the boxes of the related files are checked, click "Publish".

  

If you have an error when deploying, please check "Logs" of the application through your shinyapps.io Dashboard to figure out the cause of the error.

​