Module 04 Activity

Scenario

Turn a wide monthly return into a table that answers a question, using each verb deliberately.

What you build

A wrangling script with the row count reported at every shape change.

Steps

1. Identify any column name that is really a value and pivot to the long shape before anything else.
2. Write the pipeline with |>, one verb per line, in the order the steps happen. Break it into named objects wherever the data starts meaning something new.
3. Use select and filter to shrink first, mutate to derive, then group_by and summarise to change the grain — printing the row count either side of the summarise.
4. Band a numeric column with case_when(), testing missingness first and supplying .default. Verify one row per band by hand.
5. Before any join, check the key with anyDuplicated() and choose the join type deliberately. Compare row counts before and after.
6. Pivot back to wide only for the final table a person will read.

Evidence to hand in

• The long-form table and what column name was really a value.
• The pipeline, one verb per line.
• Row counts either side of the summarise.
• The banded column, with one verified row per band.
• The key check and the row counts around the join.
• The final wide table.

Review checklist

• No column name in the working data is a date, a category or a year.
• .groups = 'drop' is used, or the grouping is dropped deliberately.
• The case_when() tests missingness first and ends with .default.
• anyDuplicated() was run on the join key before joining.
• Row counts are reported around both the summarise and the join.

Mark complete