Appendix C — Data sources

Every dataset used in this book, where it came from, and what you need to do to get hold of it. Each is also credited at the point it first appears; this is the single list.

C.1 peopleanalyticsdata

Most of the book runs on the peopleanalyticsdata package, created by Keith McNulty. MIT licensed, and on CRAN:

install.packages("peopleanalyticsdata")
library(peopleanalyticsdata)

data(package = "peopleanalyticsdata")   # everything available

Dataset	Shape	Used in
salespeople	351 rows — sales, customer rating, performance tier, promoted	1–9, 18, 20
employee_survey	2,833 rows — four-point Likert items in blocks	1–3, 17, 18
managers	571 rows — test score, city, tenure, transfers, team size	8, 12, 13, 14
graduates	173 rows — median salary and unemployment by major	15
promotion	1,134 rows — time to promotion, flexible working, diversity programme	16, 19
recruiting	966 rows — hiring decisions, test scores, three interviewer ratings	exercises
learning	4,974 rows — repeated feedback from 326 participants	12 (crossed effects)

C.2 Downloaded separately

turnover — real, anonymised company data originally shared by Edward Babushkin and redistributed on Kaggle under CC BY-NC-SA 4.0. Genuine tenure in months, right-censored for those still employed, which is why Chapter 17 uses it rather than a constructed alternative.

Not bundled with this book, because the licence is share-alike and the data is not ours to redistribute. Download it from https://www.kaggle.com/datasets/davinwijaya/employee-turnover and save it as data/turnover.csv before running Chapter 17. See data/README.md in the repository for the detail.

C.3 Simulated data

Chapter 22 simulates its own panel, because no public People Analytics dataset has both a time dimension and a policy change. That is stated in the chapter and is deliberate: when you generate the data you know the true answer, which is the only situation in which you can check whether a method recovers it.