Scottish Climate Survey Dashboard

A Shiny application for analyzing weighted survey data with grouping and statistical comparisons.

Project Structure

R/
  ├── app.R                    # Main application entry point
  ├── global.R                 # Global settings and library loading
  ├── ui.R                     # User interface definition
  ├── server.R                 # Server logic
  ├── functions/               # Modular function files
  │   ├── utils.R              # General utility functions
  │   ├── data_io.R            # Data reading and validation
  │   ├── factors.R            # Factor creation and manipulation
  │   ├── ui_helpers.R         # UI generation helpers
  │   ├── analysis.R           # Survey data analysis functions
  │   ├── plotting.R           # Plotting and visualization
  │   ├── statistics.R         # Statistical tests
  │   └── grouped_analysis.R   # Multi-select question analysis
├── tests/                   # Unit tests
│   ├── testthat.R           # Test runner
│   └── testthat/            # Individual test files
│       ├── test-utils.R
│       ├── test-factors.R
│       └── ...
└── README.md                # This file

Installation

1. Install required packages:

install.packages(c(
  "shiny", "readxl", "dplyr", "tidyr", "stringr", "forcats",
  "ggplot2", "DT", "scales", "survey", "tibble", "testthat"
))

2. Optional: Install RColorBrewer for enhanced color palettes:

install.packages("RColorBrewer")

Running the Application

# From the project root directory
shiny::runApp()

Or simply run app.R in RStudio.

Running Tests

# Run all tests
testthat::test_dir("tests/testthat")

# Or use devtools
devtools::test()

Features

1. Plots - Raw

• Weighted proportion plots for single questions
• Primary and optional secondary grouping variables
• Customizable Likert color palettes
• Exclude specific response levels (Don't know, Prefer not to say, etc.)
• Download plots as JPG and tables as CSV

2. Plots - Grouped

• Analysis of multi-select questions (@ prefixed variables)
• Heatmap visualization of selection rates by group
• Significance testing with Rao-Scott chi-squared tests
• Significance markers (★) for notable differences

3. Statistics

• Chi-squared tests for overall differences across groups
• Pairwise Wald comparisons with Bonferroni correction
• Custom grouping of response levels
• Custom grouping of demographic groups

Data Format

The application expects an Excel workbook (.xlsx) with three sheets:

Data

Contains the raw survey responses with one row per respondent.

Variable Labels

Two columns:

• Variable: Variable name (must match Data columns)
• Label: Human-readable label

Value Labels

Three columns:

• Variable: Variable name
• Value: Coded value (e.g., "1", "2")
• Label: Human-readable label for the value

Function Organization

Functions are organized by purpose for easy testing and maintenance:

• utils.R: General utilities (null coalescing, filename sanitization)
• data_io.R: Reading and validating Excel workbooks
• factors.R: Creating factors with value labels, collapsing levels
• ui_helpers.R: Building UI elements (dropdowns, headers)
• analysis.R: Preparing data for analysis, creating survey designs
• plotting.R: Creating plots and tables
• statistics.R: Statistical tests (chi-squared, pairwise comparisons)
• grouped_analysis.R: Multi-select question analysis and visualization

Testing

Unit tests are provided for core functions. Tests are organized by function file:

• test-utils.R: Tests for utility functions
• test-factors.R: Tests for factor manipulation
• Add more test files as needed for comprehensive coverage

To add new tests:

1. Create a test file in tests/testthat/ named test-[function_file].R
2. Follow the naming convention: test_that("description", { ... })
3. Run tests with testthat::test_dir("tests/testthat")

Development Notes

• Survey designs use the survey package for weighted analysis
• Negative codes in numeric variables are treated as missing
• The app supports files up to 100 MB
• Lonely PSUs are adjusted automatically in survey designs