Compares a model with a coarser version of itself, column by column, and says which quantities survived. The comparison is by class, slot and column, not by object: aggregation merges corridors and drops the ones internal to a coarse region, so a per-object diff would read that as loss.
Arguments
- fine
the model before aggregation.
- coarse
the model
aggregate_model_regions()returned.- tol
relative tolerance for an extensive total (default
1e-6).
Value
A data.frame with one row per class, slot and column: the rule,
the fine and coarse totals, the absolute and relative difference, the two
ranges and a verdict. Printed as a short summary with the exceptions
listed.
Details
Each column is judged by its own aggregation rule, read from the same table the aggregation used:
sum– an extensive quantity (a stock, a capacity, a demand, an availability): its total must be unchanged, and any difference is reported with its relative size.weighted_mean,mean,copy– a rate, a price or an invariant: a weighted mean cannot leave the range of the values it came from, so the verdict is whether the coarse range still sits inside the fine one.
A trade row whose object count fell is reported as by construction
rather than as a difference: a corridor between two regions of one coarse
region is internal to it and is dropped, so its capacity is not expected to
survive. That is the aggregation's meaning, not an error – which is exactly
why it is named rather than hidden.
See also
aggregate_model_regions(), check_model_levels(),
verify_solution()
Other validation:
check_model_levels()
