Refuse a cohort that is not the one the model was developed on
Source:R/modeling_rms.R
mm_same_development.RdA row count cannot distinguish a cohort from a different cohort of the same
size, and every basis mm_validate() reports describes the development
sample. Scoring a different one and calling the result apparent is not a
smaller error than scoring none.
Arguments
- model
A
mm_model().- dev
The development rows a caller has re-derived.
- what
What the caller is about to do, for the message.
Details
The comparison is order sensitive, because the pairing of predictors to outcome is what a c-statistic measures: the same rows in a different order are the same sample, but nothing cheap tells that apart from the same predictor rows carrying permuted outcomes, which is a different sample entirely. So a reordered cohort is refused too, and the message says so.