How to use treeModule

Introduction

soroban’s treeModule generates Decision tree.

In this article, we’ll use TeachingRatings dataset of AER

library(AER)
data('TeachingRatings')

This article is based on 0.0.1 Version of soroban

Declare module

soroban’s module assumes that used in the Shiny application.

and You can use snippet(type shinyapp) to build very basic shiny application.

library(shiny)

ui <- fluidPage(
 
)

server <- function(input, output, session) {
  
}

shinyApp(ui, server)

This application will show nothing.

So let’s add treeModule in ui.

ui <- fluidPage(
  mod_treeModule_ui(
    id = 'module'
  )
)

also, treeModule in server.

server <- function(input, output, session) {
  mod_treeModule_server(
    id = 'module', 
    inputData = reactive(TeachingRatings)
  )
}

So final (which is very basic) code will like this. (Assume data from AER loaded.)

library(shiny)

ui <- fluidPage(
  mod_treeModule_ui(
    id = 'module'
  )
)

server <- function(input, output, session) {
  mod_treeModule_server(
    id = 'module', 
    inputData = reactive(TeachingRatings)
  )
}

shinyApp(ui, server)

You should notice 2 things.

1. both id in ui and server should be same.
2. inputData in server should be format of reactive

Structure of treeModule

treeModule is consisted with Control Area and Result Area

and below using flow.

1. Declare module (we did already)
2. select variable to generate tree. 2-1. (Optional) adjust tree layout option like Edge and Node, Terminal of tree
3. build Tree

Usage of treeModule

Using TeachingRatings, we’ll see which factor effects teacher’s evaluation.

Set X as beauty, minority, age ,gender,division,native,tenure (order is not important)

and Y as eval.

that is we want to model Evaluation with 7 factors

Also, the treeModule supports regression too.

select nodePlot options to beauty, gender, minority (typo will fixed in further version)

and after Tree, generated tree will be shown.

Result shows that, beauty and other factor’s effect to teacher evaluation. and below scatter plot is regression for relation between beauty and eval based on gender.

(We consider change nodePlot option may not required in further version)

However if we change some character type value to factor, (which is more accurate) result will also changed.

For any issue or suggestion, please make issue in soroban’s github.