RAnEnExtra::verifySpreadSkill calculates the spread skill correlation between ensemble variance and squared error. Read details for caveats.

verifySpreadSkill(
  anen.ver,
  obs.ver,
  boot = F,
  R = 1000,
  na.rm = T,
  keep.cor = F
)

Arguments

anen.ver

A 4-dimensional array. This array is usually created from the value column of the analogs member in the results of RAnEn::generateAnalogs. The dimensions should be [stations, times, lead times, members].

obs.ver

A 3-dimensional array. The dimensions should be [stations, times, lead times]. You can generate the array using RAnEn::alignObservations.

boot

Whether to use bootstrap.

R

The number of bootstrap replicates. Used by the function boot::boot.

na.rm

Whether to remove NA values.

keep.cor

Whether to keep all spread and errors.

Details

Spread skill correlation without normalization or binning has several caveats during its interpretation. Please make sure you have read through several materials before applying this verification technique. Alternatively, RAnEn::verifyBinnedSpreadSkill is an improved version of this function.

Bootstrap confidence interval is defaulted to 0.95. To change this, use options(RAnEnExtra_boot_conf = 0.9).