Summarise a cross-check into per-parameter metrics

One row per term and level: the mean and root mean square of estimate - reference, the share of fit intervals covering the reference, the share of rows whose intervals overlap, and the correlation and Lin's concordance between the two sides.

Usage

# S3 method for class 'bmmtools_cross_check'
summary(object, ...)

# S3 method for class 'bmmtools_cross_check_summary'
summary(object, ...)

Arguments

object

A bmmtools_cross_check object from cross_check().

...

Not used.

Value

A bmmtools_cross_check_summary tibble with the columns term, level, scale, n, n_converged, bias, rmse, coverage, share_overlap, r, r_low, r_high, ccc, ccc_low and ccc_high.

Details

bias here is a difference from a comparison value, not from a known truth; see cross_check().

r and ccc come from the pairs of estimate and reference in the group and are NA, never 0, below three complete pairs or without spread on either side. At the population level with one fit each term contributes a single pair, so both are NA and the table is the per-term difference and coverage; at the subject level they are the comparison across subjects. share_overlap is NA for a reference without intervals. n_converged counts the rows whose fit passed check_convergence() and is NA when no row carries a verdict.