sens.var.samp: Variance-based sensitivities of a deterministic model

In baccione-eawag/EawagSchoolTools: Miscellaneous Tools for Eawag Summer School in Environmental Systems Analysis

Description

Function to calculate the first order variance-based sensitivity coefficients from a parameter sample and the corresponding result sample.

Usage

sens.var.samp(parsamp,ressamp,nbin=NA,method="smooth",order=2,
              bandwidth=1,span=0.75,sd.rel=0.1,plot=F)

Arguments

parsamp	matrix containing the parameter sample (each row corresponds to a sample point)
ressamp	vector (of only one result per parameter sample point) or matrix of results corresponding to the parameter sample (each row provides the results corresponding to the parameter values in the same row of parsamp)
nbin	number of quantile intervals for which the conditional means are calculated, default is the square root of the sample size
method	"smooth", "loess", "glkerns", "lokerns", "lpepa", "lpridge": routine to be used for smoothing
order	order of local regression polynomial or of kernel
bandwidth	method="lpepa" or method="lpdidge" only: bandwidth of the smoothing algorithm
span	method="loess" only: fraction of points used for local regression
sd.rel	method="smooth" only: standard deviation of normal distribution of smoothing algorithm relative to the 99% quantile interval
plot	logical variable indicating if a scatter plot of the relationship between parameter and model output should be plotted

Value

Returns a list with two elements:

var	vector with the total variance of each column (each model output) of the ressamp matrix
var.cond.E	matrix with the variance of the conditional expected value of each parameter (columns) and each model output (rows)

Author(s)

Peter Reichert <[e-mail]>

baccione-eawag/EawagSchoolTools documentation built on Dec. 19, 2021, 6:38 a.m.