Survey Crosstab Explorer

Project roadmap

1. 01

   Set up the project and sample data

   ~1h

   Install R and RStudio, then install the packages shiny, dplyr, tidyr, ggplot2, readr, infer and rmarkdown. Create app.R with a minimal Shiny app (ui, server and shinyApp) that shows a title. Add a sample survey CSV in data/ with at least 100 rows and 4 categorical columns, such as age_group, region, satisfaction and would_recommend. Done when: running the app opens a page with the title, and the sample file is in the repo.

2. 02

   Upload a CSV and pick two questions

   ~1.5h

   Add a fileInput for the CSV, read it with readr::read_csv, and fill two selectInput dropdowns with its column names: one for rows and one for columns. Show a clear message if the file can't be read or has fewer than two columns. Until a file is uploaded, use the sample data. Done when: uploading a file updates both dropdowns with its column names.

3. 03

   Build the crosstab

   ~2h

   When two different columns are selected, count each combination of answers with dplyr::count and reshape the result with tidyr::pivot_wider into a table: one row per answer to the first question, one column per answer to the second, containing counts. Add a second table with row percentages rounded to one decimal. Treat missing answers as a category called (missing). Done when: the app shows both tables for the selected pair.

4. 04

   Add a chart

   ~1.5h

   Draw a stacked bar chart with ggplot2: one bar per answer to the first question, split by the answers to the second, using row percentages so every bar reaches 100%. Add axis labels and a title naming both questions. Done when: the chart updates whenever either dropdown changes.

5. 05

   Test whether the answers are related

   ~1.5h

   Use infer to run a chi-square test of independence between the two questions, for example with chisq_test. Show the statistic, the p-value, and a plain-language sentence: likely related if p < 0.05, no clear relationship otherwise. If any expected count is below 5, add a warning that the result may be unreliable. Done when: the test result and sentence appear below the tables.

6. 06

   Export a report

   ~1.5h

   Add a Download report button using downloadHandler that renders an R Markdown file (report.Rmd) to HTML, containing the two chosen questions, both tables, the chart and the test result. Done when: clicking the button downloads an HTML report that matches what's on screen.

Resources

• DocsMastering Shiny (free online book)
• DocsR for Data Science (free online book)
• Docsinfer package documentation
• Docsggplot2 documentation
• DocsR Markdown documentation