orthoEnsemble: Conditional-sampling ensemble subject to constraints
In garciapintado/rDAF: An R Data Assimilation Framework

Description Usage Arguments Details Value Author(s)

Given a background mean and standard deviation for a number of variables, the function returns random samples from the background conditional distribution, with a given inflation factor and possibly subject to fixed-value and cross-variable constraints. See details below.

1	orthoEnsemble(p, MC, m, MCcons = NULL, sdfac = 1)

`p`	LIST with named input variables as components. Each component is then a vector of length two, containing its background mean and standard deviation
`MC`	Dataframe with metadata, where the `MC[,'ispar']` is a LOGICAL column indicating which variables are to be included in the output matrix, out of all those available in `p`. Output variables are those with both `MC[,'ispar']` and `MC[,'flag']` set to `TRUE`, with length `ntheta`. `MC` has the additional 'max' and 'min' columns indicating fix-valued contraints on output samples. `MC[,'dis']` indicates the random distribution type. Available values are 'rnorm', 'truncnorm' and 'rlnorm'.
`m`	number of members in the ensemble, m has to be an exact multiplier of ntheta
`MCcons`	CHARACTER, 3-column matrix of cross-variable constraints. For each row, the first and third columns are variable names matching those in `p`, and the second column is one character in ['<','=','>']
`sdfac`	REAL, scalar or [ntheta] vector. Multiplicative factor of the standard deviation for the perturbations

Makes a conditional-sampling parameter Ensemble given the parameters in p and MC. This ensemble includes the control run [or background] as the component $b. Then one list member for each perturbed parameter. MCcons gives a number of equality and inequality constrains with one row per constraint. As it might happen that a background value violates a constrain, these are first evaluated on the background, then on the generated ensemble

A LIST with the components

`b`	[ntheta] background vector
`thetaNam`	[ntheta] name of the output variables
`dtheta`	[ntheta,mperpar] matrix of perturbations, where mperpar is the number of perturbations per thetaNam variable

Then for each thetaNam, there is an additional component in the LIST with a vector or matrix (if more than one sample is requested for each thetaNam variable) where only thetaNam is perturbed

Javier Garcia-Pintado

garciapintado/rDAF documentation built on May 25, 2019, 7:26 p.m.