weightVectorToMatrix: Convert the weights vector to a matrix

In SWS-Methodology/faoswsImputation: Package to perform ensemble imputation for various FAO domains

Description

The default weights vector assumes that all models are valid for all imputation points. However, some imputations may be extrapolations that are well outside the range of the data, and in this case not all models will be valid. Thus, this function takes the weights vector and adjusts it for such observations.

Usage

weightVectorToMatrix(x, w, modelExtrapolationRange)

Arguments

x	A numeric vector to be imputed.
w	The vector of weights, computed based on the errors of the models.
modelExtrapolationRange	A numeric vector specifying the valid range of extrapolation for each model. This vector must be the same length as w, as it's i-th element gives the extrapolation range for the i-th weight. Note: care should be taken to ensure the weight vector's i-th element and the modelExtrapolationRange vector's i-th element both correspond to the same model in the ensemble.

Details

Note that the extrapolation range was not a parameter in previous versions of the code. Thus, if that element is not present, the model assumes the range is infinite. This ensures backwards-compatibility.

Value

A matrix of weights of dimension Txm, where T=number of time steps and m=number of models. The (i,j) element of this matrix, then, is the weight that the j-th model should receive at the i-th time step.

SWS-Methodology/faoswsImputation documentation built on May 9, 2019, 11:48 a.m.