nll_1taper: First Tapering Implementation
In coatless/pims_bigdata: Packaged spatial methods

Description Usage Arguments Details Value Functions Examples

The function implements the first tapering method described in the Covariance Tapering paper.

1
2
3

nll_1taper(x, n, good_dists, taps, ia, ja, z, rescol)

nll_1taper_parallel(x, n, good_dists, taps, ia, ja, z, rescol, cores)

`x`	A `double`
`n`	A `double` that indicates the matrix row/col.
`good_dists`	A `VectorXd` that contains the distances which meet tapering requirements
`taps`	A `VectorXd` results of applying the taper function
`ia`	A `VectorXd` containing the ordered row indices
`ja`	A `VectorXd` containing the ordered column indices
`z`	A `VectorXd` element position in matrix
`rescol`	An `int` that contains how many non-zero elements should be in a column.
`cores`	An `int` that indicates the number of cores to divide the task over.

Currently the EIGEN sparse matrix cholesky decomposition differs from that of SPAM. As a result, the forward solve step finds a different vector. This in turn returns a higher distval.

A double

nll_1taper_parallel:

data(anom1962)
d = rdist_earth1(loc)
setup.eigen = make_tapersetup_eigen(d,taprange = 50)
val = nll_1taper(20, setup.eigen$n,
                 setup.eigen$good.dists,
                 setup.eigen$taps,
                 setup.eigen$ia,
                 setup.eigen$ja,
                 z,
                 setup.eigen$rescol)