Return decomposition of the Brier Score into Reliability, Resolution and Uncertainty, and estimated standard deviations
Usage
BrierDecomp(p, y, bins = 10, bias.corrected = FALSE)
Arguments
p
vector of forecast probabilities
y
binary observations, y[t]=1 if an event happens at time t, and y[t]=0 otherwise
bins
binning to estimate the calibration function (see Details), default: 10
bias.corrected
logical, default=FALSE, whether the standard (biased) decomposition of Murphy (1973) should be used, or the bias-corrected decomposition of Ferro (2012)
Value
Estimators of the three components and their estimated standard deviations are returned as a 2*3 matrix.
Details
To estimate the calibration curve, the unit line is categorised into discrete bins, provided by the `bins` argument. If `bins` is a single number, it specifies the number of equidistant bins. If `bins` is a vector of values between zero and one, these values are used as the bin-breaks.
References
Murphy (1973): A New Vector Partition of the Probability Score. J. Appl. Met. 10.1175/1520-0450(1973)012<0595:ANVPOT>2.0.CO;2
Ferro and Fricker (2012): A bias-corrected decomposition of the Brier score. QJRMS. 10.1002/qj.1924
Siegert (2013): Variance estimation for Brier Score decomposition. QJRMS. 10.1002/qj.2228