Companion training and practice
Author: Jaime Yan.
Reading develops understanding; training asks you to deliver independently. The materials are organized around tasks, with data, starter scripts, acceptance checks, and a reference implementation. You do not need to download a conference workshop first.
There is currently one complete integrated lab. Every chapter already has progressive exercises and a capstone, but separate data packages and reference solutions are not yet available for all of them.
Multi-site data quality and analysis handoff
Three sites have submitted simulated quality-score observations. Your task is to deliver a summary table, a chart, and a record of how each input row was handled, so that a recipient can reproduce the outputs in a fresh R session.
The task covers exact duplicates, conflicting duplicates, missing scores, out-of-range values, and empty sites. Specify the rule for each case and explain the denominator used for each mean. A site can represent a business location; research analysts can also use the exercise to practice multi-site data intake and quality checks.
The records are artificially constructed business-quality observations. They contain no patient records and represent neither clinical endpoints nor treatment effects.
Download the complete English lab Read the task and rubric online
| Item | Details |
|---|---|
| Prerequisites | CSV files, data frames, functions, and basic plotting; R ≥ 4.1; no additional packages |
| Suggested time | 90–120 minutes, optionally split into two sessions |
| Included files | Simulated CSV, starter.R, solution.R, check.R, and the full task guide |
| Sequence | Predict and inspect the data, implement functions, then generate the deliverables |
| Acceptance | Input contract, duplicate and missing-value handling, boundaries, summary denominators, and export read-back |
| Deliverables | Source code, summary CSV, audit CSV, PNG chart, and a one-page handoff note |
Getting started
Extract the download and open a terminal in training/analysis-handoff/. Read the README first. For the first 60 minutes, work without AI and keep the reference solution closed.
Rscript starter.RSave your implementations of the four functions as learner.R, then run the acceptance checks:
Rscript check.R output-check learner.RAfter an independent attempt, run the reference implementation and compare the designs:
Rscript solution.R output-reference
Rscript check.R output-reference-checkWhen a check fails, identify the violated data contract, write a minimal example, and then correct the function. After the checks pass, review the chart labels and handoff note to ensure the recipient can understand the data’s limitations.
Connecting practice to the chapters
| Question from the lab | Return to |
|---|---|
| Which arguments does a function need, and how do you reduce hidden dependencies? | 1.1 Writing functions |
| How can the same process handle a second batch? | 1.2 Iteration |
| How should observed counts and missingness be displayed? | 1.6 Tables, 1.8 Plots |
| How should boundary inputs be checked? | 3.2 Unit testing |
| How will the recipient know the data source and execution conditions? | 2.8 Reports, 4.8 Environments and reproducibility |
For teaching, use the AI-off period for independent work, then arrange a peer handoff. The recipient should use only the README and outputs, attempt a rerun, and identify one remaining uncertainty. Use the reference solution to compare approaches; identical code is not the grading criterion.