4  Task Two - Data Cleaning

4.1 Objective: Practice clean a messy dataset

This week’s assignment is to use simple dplyr commands to take a messy dataset, understand it and clean it.

ImportantUsing AI

• For this exercise we are asking you Not to use LLMs such as Copilot or ChatGPT.

• This is because this is your opportunity to build familiarity with the process of thinking about your data and learning basic R functions

• It is tempting to dump code into AI and ask for a fix - but please hold off it will benefit you to try and understand it for yourself

4.1.1 Step 1: Download the Scripts

• Access Blackboard → Week 2 → “Cleaning Data”

• Download the R scripts to your local machine.

4.1.2 Step 2: Upload to Posit Cloud

• Open your script in a NEW project.

• Upload the downloaded scripts into the project.

Think organised project!

4.1.3 Step 3: Checklist

4.1.3.1 Set up:

• Load required packages (tidyverse, janitor, lubridate, stringr, forcats, skimr).

• Import data (read_csv / here()).

• (Optional) Immediately standardise names:

  • Apply snake_case: janitor::clean_names().
  • Manually rename awkward or long columns: dplyr::rename().
  • Ensure no spaces, punctuation, or case inconsistencies that force backticks.
  • use sensible and clear names

4.1.3.2 Inspect structure & quality:

• Dimensions & glimpse: dim(), glimpse(), head(), tail().

• Types & summaries: str(), summary(), skimr::skim().

• Check for unexpected types (e.g. dates as character, numbers as character).

• Confirm data is (or can be made) tidy{Appendix C}: one row = one observation, one column = one variable.

4.1.3.3 Standardise values (within columns)

• Collapse verbose categorical labels (case_when(), if_else()).

• Harmonise capitalisation (stringr::str_to_title / str_to_lower).

• Trim stray whitespace (stringr::str_trim()).

• Make any required new variables with mutate()

• Handle dates parse with correct order (lubridate::dmy / mdy / ymd / etc.).

4.1.3.4 Check duplicates & integrity

• Whole-row duplications

• Detect typos or impossible values

4.1.3.5 Factors & ordering

• Convert categorical text to factors: mutate(across(cols, forcats::as_factor)).

• Set meaningful order: fct_relevel() (e.g. smol < mid < chonk).

• (Optional) Create classification / grouping variables

4.1.3.6 Summarise & validate

• Re-run skim() / summary() after transformations.

• Spot remaining NAs; decide whether to keep or drop.

• Sanity-check ranges (e.g. min/max dates, value distributions).

4.1.3.7 Readable code

• Use pipes for readability

• Assign new R objects only at meaningful checkpoints - use clear names.

• Keep all steps in a script (e.g. 01_import_penguins_data.R).

• Ensure script runs cleanly from a fresh R session (no hidden state).

• Add brief comments explaining each transformation.

• Save cleaned object (e.g. write_csv() or saveRDS()).

• Do not overwrite or edit the raw data file.

4.1.4 Step 4: Submission / Discussion

1. Submit your fixed scripts to Blackboard.