attribute {verification}

Attribute plot

Description

An attribute plot illustrates the reliability, resolution and uncertainty of a forecast with respect to the observation. The frequency of binned forecast probabilities are plotted against proportions of binned observations. A perfect forecast would be indicated by a line plotted along the 1:1 line. Uncertainty is described as the vertical distance between this point and the 1:1 line. The relative frequency for each forecast value is displayed in parenthesis.

Usage

## Default S3 method:
attribute(x, obar.i,  prob.y = NULL, obar = NULL,
    class = "none", main = NULL, CI = FALSE, n.boot = 100, alpha = 0.05,
    tck = 0.01, freq = TRUE, pred = NULL, obs = NULL, thres = thres,
    bins = FALSE, ...)

## S3 method for class 'prob.bin'
attribute(x, ...)
       

Arguments

x

A vector of forecast probabilities or a “prob.bin” class object produced by the verify function.

obar.i

A vector of observed relative frequency of forecast bins.

prob.y

Relative frequency of forecasts of forecast bins.

obar

Climatological or sample mean of observed events.

class

Class of object. If prob.bin, the function will use the data to estimate confidence intervals.

main

Plot title.

CI

Confidence Intervals. This is only an option if the data is accessible by using the verify command first. Calculated by bootstrapping the observations and prediction, then calculating PODy and PODn values.

n.boot

Number of bootstrap samples.

alpha

Confidence interval. By default = 0.05

tck

Tick width on confidence interval whiskers.

freq

Should the frequecies be plotted. Default = TRUE

pred

Required to create confidence intervals

obs

Required to create confidence intervals

thres

thresholds used to create bins for plotting confidence intervals.

bins

Should probabilities be binned or treated as unique predictions?

...

Graphical parameters

Note

Points and bins are plotted at the mid-point of bins. This can create distorted graphs if forecasts are created at irregular intervals.

Author(s)

Matt Pocernich

References

Hsu, W. R., and A. H. Murphy, 1986: The attributes diagram: A geometrical framework for assessing the quality of probability forecasts. Int. J. Forecasting 2, 285–293.

Wilks, D. S. (2005) Statistical Methods in the Atmospheric Sciences Chapter 7, San Diego: Academic Press.

See Also

verify reliability.plot

Examples

## Data from Wilks, table 7.3 page 246.
 y.i   <- c(0,0.05, seq(0.1, 1, 0.1))
 obar.i <- c(0.006, 0.019, 0.059, 0.15, 0.277, 0.377, 0.511, 
             0.587, 0.723, 0.779, 0.934, 0.933)
 prob.y<- c(0.4112, 0.0671, 0.1833, 0.0986, 0.0616, 0.0366,
            0.0303,  0.0275, 0.245, 0.022, 0.017, 0.203) 
 obar<- 0.162
 
attribute(y.i, obar.i, prob.y, obar, main = "Sample Attribute Plot")  

## Function will work with a ``prob.bin'' class objects as well.
## Note this is a random forecast.
obs<- round(runif(100))
pred<- runif(100)

A<- verify(obs, pred, frcst.type = "prob", obs.type = "binary")
attribute(A, main = "Alternative plot", xlab = "Alternate x label" )
## to add a line from another model
obs<- round(runif(100))
pred<- runif(100)

B<- verify(obs, pred, frcst.type = "prob", obs.type = "binary")
lines.attrib(B, col = "green")


## Same with confidence intervals
attribute(A, main = "Alternative plot", xlab = "Alternate x label", CI =
TRUE)

#### add lines to plot
data(pop)
d <- pop.convert()
## internal function used to
## make binary observations for
## the pop figure.

### note the use of bins = FALSE
mod24 <- verify(d$obs_rain, d$p24_rain,
    bins = FALSE)

mod48 <- verify(d$obs_rain, d$p48_rain,
    bins = FALSE)
plot(mod24, freq = FALSE)

lines.attrib(mod48, col = "green",
    lwd = 2, type = "b")