Datasets
Dataset strategy
The course uses a small number of connected datasets. Reusing the same variables across lessons helps students focus on the method rather than relearning a new context every week.
All core datasets are synthetic. They contain no real employee, customer, hotel, organizational, or student information.
1. Employee and organizational practices survey
File: data/employee_survey.csv
Rows: 480 synthetic employees
Uses: cleaning, visualization, correlations, t-tests, reliability, scale construction, exploratory factor analysis, chi-square tests, moderation, and mediation.
Key constructs include leadership, engagement, role clarity, workload, organizational support, satisfaction, performance, burnout, and turnover intention.
2. Management-training experiment
File: data/training_experiment.csv
Rows: 220 synthetic participants
Uses: independent-samples and paired-samples t-tests.
The dataset contains pre-intervention and post-intervention scores for a training and control group.
3. Marketing mix
File: data/marketing_mix.csv
Rows: 156 weekly observations
Uses: multiple linear regression, multicollinearity, residual diagnostics, and influential observations.
The variables include spending on search, social, and email advertising, discounts, a competitor price index, and weekly sales.
4. Hotel upgrades
File: data/hotel_upgrades.csv
Rows: 900 synthetic stays
Uses: logistic regression, odds ratios, and predicted probabilities.
The outcome is whether a guest received an upgrade. Predictors include loyalty status, occupancy, prior stays, spending, events, and booking channel.
5. Replication sample
File: data/employee_replication.csv
Rows: 360 synthetic employees
Uses: reproducing an analysis in a second sample and comparing estimates.
Data dictionary
A variable-level dictionary is available here:
Why synthetic data?
Synthetic data provides three teaching advantages:
- the examples can be shared publicly without privacy concerns;
- the same underlying relationships can support several methods;
- the generation process can be inspected and reproduced.
The generation scripts are available in data-raw/.
Synthetic data can look realistic but does not establish evidence about real organizations. Results from these files must never be presented as empirical findings about actual employees, firms, or customers.