When applying a scoring rule via score()
, the names of the scoring rules
become column names of the
resulting data.table. In addition, an attribute metrics
will be
added to the output, holding the names of the scores as a vector.
This is done so that functions like get_forecast_unit()
or
summarise_scores()
can still identify which columns are part of the
forecast unit and which hold a score.
get_metrics()
accesses and returns the metrics
attribute. If there is no
attribute, the function will return NULL
(or, if error = TRUE
will
produce an error instead). In addition, it checks the column names of the
input for consistency with the data stored in the metrics
attribute.
Handling a missing or inconsistent metrics
attribute:
If the metrics attribute is missing or is not consistent with the column names of the data.table, you can either
run
score()
again, specifying names for the scoring rules manually, oradd/update the attribute manually using
attr(scores, "metrics") <- c("names", "of", "your", "scores")
(the order does not matter).
Usage
# S3 method for class 'scores'
get_metrics(x, error = FALSE, ...)
Arguments
- x
A
scores
object, (a data.table with an attributemetrics
as produced byscore()
).- error
Throw an error if there is no attribute called
metrics
? Default is FALSE.- ...
unused
See also
Other get_metrics functions:
get_metrics()
,
get_metrics.forecast_binary()
,
get_metrics.forecast_nominal()
,
get_metrics.forecast_point()
,
get_metrics.forecast_quantile()
,
get_metrics.forecast_sample()