add_metrics1_d()
andadd_metrics1_c()
extracts the lpd and RPS from the Stanfit objectadd_metrics2_d()
andadd_metrics2_c()
calculates the lpd and (C)RPS from the empirical pmfThe metrics in
add_metrics2_c()
and the CRPS ofadd_metrics1_c()
are calculated using thescoringRules
package.
Usage
add_metrics1_d(df, fit)
add_metrics1_c(df, fit)
add_metrics2_d(df, support, add_samples = support)
add_metrics2_c(df, add_samples = NULL, bw = NULL)
Arguments
- df
Dataframe to add the metrics to
For
add_metrics1_c()
, it must contain a column "Score".For
add_metrics2_c()
andadd_metrics2_d()
, it must contain the columns "Samples" and "Score".
- fit
Stanfit object with parameters "lpd", and for
add_metrics1_d()
"cum_err".- support
Support of the distribution
- add_samples
Numeric vector used to initialise the distribution when computing the lpd and (C)RPS. For example, this can be used to add a uniform distribution to the vector of samples, to avoid problems at the tail of the distribution. If
NULL
, the empirical pmf is not changed. Default to the uniform distribution (i.e.support
) foradd_metrics2_d()
andNULL
foradd_metrics2_c()
. The column "Samples" is not modified whenadd_samples
is not NULL.- bw
Bandwidth, for calculating lpd, see
scoringRules::logs_sample()
. Useful to set the "resolution" of the distribution.