oscarli00 / cat

Final year honours research project implementing a multistage testing platform. Built with Spring Boot, R, MongoDB, and React.

Project 52: Computer Adaptive Testing in Educational Assessment 🐱

Computer adaptive testing (CAT) is a form of testing which uses algorithms to adjust its difficulty based on the ability of the user with the aim of providing more precise ability measurements. This project focuses on a form of CAT known as multistage testing (MST) and aims to compare the effectiveness of MST against traditional fixed length tests by implementing an online testing platform supporting both testing paradigms. The R package, mstR, is used to implement the features of the multistage test (e.g. ability estimation and module selection) which is integrated with the backend via RCaller.

Compendium resources

doc contains our compendium resources (project design, development, and testing process, and our honours dissertations!)

Backend

src/main/java/com/p4p/cat contains the backend created in Spring Boot (Java), R, MongoDB

Frontend

src/main/js contains the frontend created in React

Test questions

src/main/scripts/all.json contains the questions used to create the tests, from the Trends in International Mathematics and Science Study (TIMSS) 2011 Grade 8 mathematics assessment, converted into json format.

---

Table of Contents

• Project 52: Computer Adaptive Testing in Educational Assessment
  • Project Setup
  • Build Process
  • Deployment
• Features
  • Brief Overview of Features
  • Screenshots
    • Login Screen
    • Information Screen
    • Test Screen
  • Backend API Endpoints
    • Users
    • Questions
    • Multistage Adaptive Test (MST-CAT)
    • Fixed Length Test

---

Project Setup

1. Install R
2. Install the mstR package by entering install.packages("mstR") in the R console
3. Install the Gradle dependencies specified in the build.gradle file
4. Install frontend dependencies by running npm install in src/main/js

Note: The application uses MongoDB therefore you should set up an application.properties file in src\main\resources with a line specifying the database URI i.e. spring.data.mongodb.uri=YOUR_MONGODB_URI

Build Process

1. Build the frontend using npm run build
2. Move frontend build files to the backend static folder (src/main/resources)
3. Build the backend into a jar using gradle build

Deployment

The application is deployed on an AWS EC2 instance setup with Java, R and mstR. The database is hosted remotely using MongoDB Atlas Cluster.

Features

Brief Overview of Features

1. Administers both fixed length test (FL) and multistage test (MST)
2. FL has 30 questions, 45 minutes.
3. Our MST design is 2 stages. Every candidate receives the same testlet in the first stage. Based on their result, they are routed to one of 3 testlets (easy, medium, difficult) in the second stage.
4. All MST testlets have 10 questions, 15 minutes
5. Leverages mstR R package for MST test path.
6. We use MR2JS(MongoDB, React, R, Java, Spring Boot) stack. You can pronounce it as 'Mr 2 Js'
7. It has crash/reload recovery, so your progress or login status won't be lost
8. Made without UI frameworks, so the UI design is quite simple and barebone

Screenshots

Login Screen

This is the login screen. The users created using POST /users will have id to access the application. The other routes cannot be accessed before logging in.

Information Screen

This is the information screen, that is used to instruct the user which test they are sitting (either MST or FL tests) or inform the user that the test has ended.

Test Screen

Test screen shows timer, questions list, question, and a submit button. When the timer runs out, the test is automatically submitted. The answered questions are indicated as green on the left, and the selected answer is shown with black border.

The platform supports multiple choice questions (as shown above) and short answer questions (as shown below). Both types support images and text for the question content, and multiple choice questions can have both images and text too.

Finally, when the user submits the test, they are given a confirmation dialog informing that they cannot revisit the questions once they submit. When they confirm, the user responses are submitted to POST /mst (if MST test) or POST /fl (if FL test). The user can submit before answering all the questions.

Backend API Endpoints

We only briefly describe the available endpoints here. For the required payloads and request/response data types, refer to the backend codes and API and DB Schema.docx

Users

The following endpoints manage CRUD operations on Users

Endpoint	Description
POST /users	Create specified number of users. Test order (Multistage Test first or Fixed Length Test first) is randomly assigned.
DELETE /users	Delete all users from the database.
GET /user/{id}	Get user document of the user id {id}]

Questions

The following endpoints manage CRUD operations on Users

Endpoint	Description
POST /questions	Create questions using json payload
GET /questions/{id}	Get question document of question of {id}
GET /questions	Get documents of all the questions in the database
DELETE /questions	Delete all the questions

Multistage Adaptive Test (MST-CAT)

Endpoint	Description
GET /mst/{moduleId}	Get questions of the MST module of {id}
POST /mst	Submit user responses for an MST stage. Updates user score and assigns next module

Fixed Length Test

Endpoint	Description
GET /fixed	Get questions of the fixed length test
POST /fixed	Submit user responses for the fixed length test. Updates user score.