Data dictionary and design

Pig feeding and diarrhoea

Synthetic teaching data. All observations are invented, not the source paper’s raw data.

Researchers house eight pigs in each of 40 pens, arranged in five batches of eight pens. In each batch, they use a random order to assign four whole pens to each feed. Pigs stay in their assigned pen. A trained observer, who does not know which feed a pen receives, uses a predefined scoring rule to record whether each pig has any diarrhoea during seven days. Each pen contributes one count of affected pigs. A pig is counted at most once; the count is not a number of diarrhoea days or repeat episodes.

Researchers assign feed to whole pens: 20 pens per feed. They do not make 320 separate feed assignments to pigs. Pigs sharing a pen may share disease exposure, so their outcomes may be alike.

Column	Unit/type	Meaning
pen_id	ID	Ppen01–Ppen40; one row per pen
batch	category	batch01–batch05; eight pens per batch
feed	category	dry or fermented; assigned to pen
n_pigs	pigs	8 pigs monitored in each pen over seven days
n_diarrhoea	pigs	0–8 distinct pigs with any diarrhoea; numerator, not days/events

No values are missing; NA is the missing-value convention if a group adds external data. IDs are invented and contain no personal information. Rows are complete by construction; this does not imply missingness is harmless in a real experiment. No weights, outcomes or sample sizes were calibrated to published results.