combineSLP: Spread-Adjusted Linear Pool (SLP) of two Normals
In JuGross/ensAR: Autoregressive postprocessing methods for ensemble forecasts

Description Usage Arguments Details Value Author(s) References Examples

Computes the weight and scale parameters of the SLP of two normals from a rolling training period.

1
2
3

combineSLP(x, par_one = list(mean = NULL, sd = NULL),
  par_two = par_one, train = 90, weight_grid = seq(0, 1, 0.1),
  scale_grid = seq(0.6, 1.4, 0.1))

`x`	A vector of observations of a weather quantity.
`par_one`	A list with two vectors `mean` and `sd` of the same length as `x` respectively, providing parameters of the normal predictive distribution.
`par_two`	A list with two vectors `mean` and `sd` of the same length as `x` respectively, providing parameters of the second normal predictive distribution.
`train`	The length of the training period.
`weight_grid`	The possible values of the SLP weight.
`scale_grid`	The possible values of the SLP scale parameter.

For each forecast date (all dates except for the first train dates), the optimal combination of SLP weight and scale parameters is found on a grid determined by weight_grid and scale_grid, such that the corresponding (predictive distribution, observation)-pair minimizes the average CRPS with respect to the rolling training period of length train.

A data frame with the observation, predictive mean and standard deviation of the two normals, and SLP weight and scale parameter for the forecast period.

J. Gross, A. Moeller.

Gneiting T., Ranjan R. 2013. Combining predictive distributions. Electronic Journal of Statistics, 7, 1747–1782.