Two synthetic, fully generic files used as a tutorial by the "descriptive panel data"
tutorial in the process-improve package.
Ten assessors scored eighteen products
on nine sensory attributes; a companion table gives measured product covariates.
The dataset is synthetic to illustrate panel-consistency checks (the Mixed Assessor
Model and scale-use correction) and relating sensory attributes to product
covariates, including the trap of telling a genuine driver apart from a proxy or
a chance correlation.
Assessor: panelist id (J01-J10).
Product: product id (Product A-R).
site: a nuisance column included on purpose, to be ignored on import.
Nine attribute columns, scored as integers on a 0-10 scale: Aroma intensity,
Sweetness, Sourness, Bitterness, Firmness, Juiciness, Colour intensity,
Aftertaste, Liking. A few cells are left missing on purpose.
Three assessors are artificially made to misbehave
(one scores at random, one rates everything high, one uses only the middle of the
scale) so the panel-consistency step has something to find.