Forecast Value Function
Calculates the economic value of a forecast based on a cost/loss ratio.
value(obs, pred= NULL, baseline = NULL, cl = seq(0.05, 0.95, 0.05),
plot = TRUE, all = FALSE, thresholds = seq(0.05, 0.95, 0.05),
ylim = c(-0.05, 1), xlim = c(0,1), ...)obs |
A vector of binary observations or a contingency table summary of values in the form c(n11, n01, n10, n00) where in nab a = obs, b = forecast. |
pred |
A vector of probabilistic predictions. |
baseline |
Baseline or naive forecast. Typically climatology. |
cl |
Cost loss ratio. The relative value of being unprepared and taking a loss to that of un-necessarily preparing. For example, cl = 0.1 indicates it would cost \$ 1 to prevent a \$10 loss. This defaults to the sequence 0.05 to 0.95 by 0.05. |
plot |
Should a plot be created? Default is TRUE |
all |
In the case of probabilistic forecasts, should value curves for each thresholds be displayed. |
thresholds |
Thresholds considered for a probabilistic forecast. |
ylim, xlim |
Plotting options. |
... |
Options to be passed into the plotting function. |
If assigned to an object, the following values are reported.
vmax |
Maximum value |
V |
Vector of values for each cl value |
F |
Conditional false alarm rate. |
H |
Conditional hit rate |
cl |
Vector of cost loss ratios. |
s |
Base rate |
Matt Pocernich
Jolliffe, Ian and David B. Stephensen (2003) Forecast Verification: A Practioner's Guide in Atmospheric Science, Chapter 8. Wiley
## value as a contingency table
## Finley tornado data
obs<- c(28, 72, 23, 2680)
value(obs)
aa <- value(obs)
aa$Vmax # max value
## probabilistic forecast example
obs <- round(runif(100) )
pred <- runif(100)
value(obs, pred, main = "Sample Plot",
thresholds = seq(0.02, 0.98, 0.02) )
##########
data(pop)
d <- pop.convert()
value(obs = d$obs_rain, pred = d$p24_rain, all = TRUE)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.