r - How to fit model with individual measurement error in DiceKriging, or can it? -
i have set of 5 data points (x=10,20,30,40,50
, corresponding response values y
, noise
s.d. of y
). these data obtained stochastic computer experiments.
how can use dicekriging in r fit kriging model these data?
x <- seq(from=10, to=50, length=5) y <- c(-0.071476,0.17683,0.19758,0.2642,0.4962) noise <- c(0.009725,0.01432,0.03284, 0.1038, 0.1887)
examples online heterogeneous noise pre-specified coef.var
, coef.trend
, coef.theta
. unlikely can have a priori on these.
i have referred answer here. however, other references suggest adding nugget parameter lambda similar adding homogeneous noise, not "individual errors".
the use of km
noise quite simple:
model <- km(~1, data.frame(x=x), y, noise.var = noise, covtype = "matern3_2")
however, noise term make line search part of l-bfgs algorithm fail. may due fact is correlated y
, because when run following lines, works:
noice <- c(0.009725,0.01432,0.03284, 0.001, 0.1887) model <- km(~1, data.frame(x=x), y, noise.var = noise, covtype = "matern3_2")
Comments
Post a Comment