How a results directory falls short of the paper's fidelity
Source:R/paper_replication.R
paper_config_shortfalls.RdWhich scenarios a directory holds is visible in its rows, and
paper_replication_coverage() checks that. How they were simulated is
not: a directory holding 2 of the 11 methods at 1000 replicates yields
figures that look like the paper's but compare two methods at a tenth of
the Monte Carlo precision. Only the config the run was launched from can
show that, so the two checks are kept separate - this one is a property
of the whole directory rather than of any single figure.
Arguments
- run_config
A parsed
scenarios_config, orNULLif none was found.- requirements
A
paper_replication_requirements()list.