Datasets Used in This Book

Every number in this book comes from a real dataset or a clearly labeled simulated (_sim) one — never from a made-up example. This appendix lists every dataset the chapters actually use: what it is, how many rows it has, which variables matter most, where it came from, and which chapter(s) put it to work. Full documentation for each file — every variable defined, its units, its missing-data pattern, and the exact download and checksum — lives in that dataset’s codebook alongside the data itself.

11. Quick reference¶

Row counts above are exact counts of data rows (header row excluded) in each committed CSV, matching the count recorded in that file’s codebook. Details for each dataset — description, key variables, source, and license — follow below in the same order.

22. Real data¶

2.12.1 Kern County environmental burden — kern_calenviroscreen¶

One row per census tract in Kern County, scored by California’s CalEnviroScreen 4.0 tool for pollution burden and population vulnerability. The CalEnviroScreen score (ces_score) combines a Pollution Burden component (air quality, traffic, pesticide use, and more) with a Population Characteristics component (health outcomes and socioeconomic factors); ces_percentile ranks that score among all California tracts, so a higher percentile means a more pollution-burdened tract. Bakersfield-area tracts rank among the state’s most burdened, which is why this file anchors the book’s very first data-driven look at Kern County.

2.22.2 Kern County air quality — kern_airquality¶

Daily PM2.5 and ozone readings for 2023 from every EPA-reporting air-quality monitor in Kern County — 12 monitors, five of them in Bakersfield. Each row is one monitor’s reading for one pollutant on one day; the file mixes 1,554 PM2.5 rows and 2,822 ozone rows for 4,376 total. Winter temperature inversions trap PM2.5 in the southern San Joaquin Valley, producing the right-skewed pollution spikes this book uses to teach distribution shape, sampling distributions, and group comparisons.

2.32.3 Kern County oil and jobs — kern_energy_employment¶

One row per calendar year (2015–2024), joining California crude-oil production (Kern produces roughly 70% of the state’s total, though EIA reports only a statewide figure) with Kern County’s labor force and the Bakersfield metro area’s oil-sector payroll employment. It is a small, real-data panel built specifically to give the “oilfield employment” hook this course’s grant narrative calls for.

2.42.4 A national health snapshot — nhanes_subset¶

A 10,000-person teaching sample of demographics and body measurements from the CDC’s National Health and Nutrition Examination Survey (NHANES), 2009–2012 cycles, distributed through the CRAN NHANES R package. This is the book’s one non-Kern dataset — used where the teaching goal needs a large general sample (Normal-model fitting, CLT demonstrations with a big population, comparing means across sex or age).

33. Simulated classroom data (_sim)¶

These files are generated by a seeded, re-runnable script rather than downloaded. Each is built to have realistic structure — plausible distributions, a genuine treatment effect, honest missing data — while staying clearly and permanently labeled as simulated.

3.13.1 Opening-day class survey — firstday_survey_sim¶

A simulated “index card” survey of 150 students on the first day of a MATH 2200-like course at a commuter Hispanic-Serving Institution: major, class standing, commute time, work hours, sleep, study plans, and self-reported statistics anxiety. Mild, deliberately built-in relationships (more work hours associate with less sleep and less planned study time) make it useful for scatterplots, correlation, and missing-data practice — three of the optional fields carry realistic light missingness by design.

3.23.2 Kern County crop panel — kern_crops_sim¶

A simulated Kern-flavored agricultural panel: 8 commodities (almonds, pistachios, table and wine grapes, oranges, tangerines/mandarins, carrots, cotton) across 9 crop years (2015–2023), one row per commodity-year. Real USDA NASS Quick Stats data was the intended source, but the live NASS API requires a free, human-registered key that the autonomous build could not self-provision, so this file uses simulated values plausible for a top U.S. agricultural county — magnitudes only, not a claim about actual Kern production. Harvested acres, yield, production, and value are built to stay internally consistent (production = harvested_acres × yield_per_acre), which makes the file useful for teaching derived quantities as well as ordinary summaries.

3.33.3 Wearable fitness study — fitness_tracker_sim¶

A simulated randomized study: 200 participants (100 treatment, 100 control) in an 8-week activity-coaching trial, with four wearable-style daily-average measures. A modest, built-in treatment effect (more steps and active minutes, lower resting heart rate, slightly more sleep in the treatment group) makes this the book’s dataset of choice for two-sample inference — Welch’s t-test, a difference-of-means confidence interval, and effect size.