Krige/interpolate from observed nodes using a Gaussian precision matrix

Conditional mean under a Gaussian model specified by a precision matrix Q_new.

kriging_from_precision(f_obs, idx_obs, Q_new, nugget = 0, keep_obs = TRUE)

Arguments

f_obs

Numeric vector length |O| or matrix |O| x p.

idx_obs

Integer vector of observed indices (subset of 1:K_new).

Q_new

K_new x K_new precision matrix (dense or sparse).

nugget

Nonnegative scalar: add nugget*I to Q_UU for stability.

keep_obs

If TRUE, returns full length K_new (or K_new x p) with obs filled in.

Value

Full vector/matrix if keep_obs=TRUE; else list(pred_idx, f_pred).