controlBMAgamma0: Control parameters for BMA precipitation modeling

Description

Specifies a list of values controling the Bayesian Model Averaging fit of a mixture of gammas with a point mass at 0 to ensemble forecasts for precipitation.

Usage

controlBMAgamma0(maxIter = Inf, tol = sqrt(.Machine$double.eps), 
                 power = (1/3), rainobs = 10, 
                 init = list(varCoefs = NULL, weights = NULL),
             optim.control = list(ndeps = rep( sqrt(.Machine$double.eps), 2)))

Value

A list whose components are the input arguments and their assigned values.

Arguments

maxIter: An integer specifying an upper limit on the number of iterations for fitting the BMA mixture via EM. The default is Inf, which sets no upper limit on the number of iterations, so that the convergence criterion based on eps is used.
tol: A numeric convergence tolerance. The EM fit for the mixture of gammas is terminated when the relative error in successive objective values in the M-step falls below tol. The default is sqrt(.Machine$double.eps), which is approximately 1.e-8 on IEEE compliant machines.
power: A scalar value giving the power by which the data will be transformed to fit the models for the point mass at 0 and mean of nonzero observations. The default is to use the 1/3 power of the data. The untransformed forecast is used to fit the variance model.
rainobs: An integer specifying a minimum number of observations with nonzero precipitation in the training data. When necessary and possible, the training period will be extended backward in increments of days to meet the minimum requirement. It is not possible to fit the BMA model for precipitation without sufficient nonzero observations. The default minimum number is 10. It many instances fewer nonzero observations may suffice, but it could also be that more are needed to model precipitation in some datasets.
init: An optional list of initial values for variance coefficients and weights. The default is to start with the variance coefficients equal to 1, and with equal weights for each member of the ensemble.
optim.control: Control parameters for the optim function used in the M-step of EM. The default here is list(ndeps = rep( sqrt(.Machine$double.eps), 2)), which assigns a smaller finite-difference step size than the optim default of 1.e-3. To use the default control parameters for optim, set optim.control=NULL.

References

J. M. Sloughter, A. E. Raftery, T Gneiting and C. Fraley, Probabilistic quantitative precipitation forecasting using Bayesian model averaging, Monthly Weather Review 135:3209--3220, 2007.

C. Fraley, A. E. Raftery, T. Gneiting and J. M. Sloughter, ensembleBMA: An R Package for Probabilistic Ensemble Forecasting using Bayesian Model Averaging, Technical Report No. 516R, Department of Statistics, University of Washington, 2007 (revised 2010).

Examples

Run this code

  data(ensBMAtest)

  ensMemNames <- c("gfs","cmcg","eta","gasp","jma","ngps","tcwb","ukmo")
  
  obs <- paste("PCP24","obs", sep = ".")
  ens <- paste("PCP24", ensMemNames, sep = ".")

  prcpTestData <- ensembleData( forecasts = ensBMAtest[,ens],
                                dates = ensBMAtest[,"vdate"], 
                                observations = ensBMAtest[,obs], 
                                station = ensBMAtest[,"station"],
                                forecastHour = 48,
                                initializationTime = "00")

if (FALSE)  # R check
  prcpTestFit1 <- ensembleBMAgamma0( prcpTestData, trainingDays = 30,
       control = controlBMAgamma0(power = (1/4)))


# for quick run only; use more training days for forecasting
  prcpTestFit1 <- ensembleBMAgamma0( prcpTestData[1:14,], trainingDays = 6,
       control = controlBMAgamma0(power = (1/4)))

Run the code above in your browser using DataLab

The Learning Leader's Guide to AI Literacy