The honest form of the "sequential update" question: after new data (or a new model),
how much did the ranking actually move? Reports the rank correlation, the retention
in the selected fraction, and the animals that moved most, the numbers a breeder
checks before trusting an updated evaluation.
Usage
rank_drift(old, new, top = 100)
Arguments
- old
named vector of EBVs from the previous evaluation
- new
named vector of EBVs from the updated one
- top
size of the selected group for the retention count (default 100, capped
at the number of common animals)
Value
list with n_common, pearson, spearman, retention (how many of the
old top-top remain in the new top-top), top, and movers (the ten largest
rank changes, positive = climbed)
