Skip to content

MultivariateSimulation

Several correlated variables simulated through independent factors.

Every realization simulates each factor with its own seed, back-transforms the factors together at the nodes, and only then averages to blocks.

Parameters:

Name Type Description Default
transform (PCA, MAF, StepwiseConditional or PPMT)

Template, fitted afresh by fit, that turns the variables into factors. PPMT, which also makes them Gaussian and takes declustering weights, is the usual choice.

 required
simulators sequence of SGS or TurningBands

One per factor, in factor order, each with the normal-score variogram of its factor.

 required

fit method descriptor 

fit(coords, data, *, weights=None, holes=None, impute=None)

Fits the transform to the data, then each simulator to its factor. Samples missing a variable are dropped with a warning, unless impute; of samples sharing a location, the first is kept, with a warning naming their holes.

Parameters:

Name Type Description Default
coords (array_like, shape(n, 2) or (n, 3), PointSet or BlockModel)
required
data array_like, shape (n, variables), or list of str

One column per simulator, or the columns of coords holding them; NaN or null marks a missing variable.

 required
weights array_like or str

Declustering weights, for the transform when it takes them (PCA, StepwiseConditional, PPMT) and for each factor's normal scores.

 None
holes array_like or str

Drill-hole ids or names, for max_per_hole and the warning on samples sharing a location.

 None
impute bool or GaussianImputer

Keep samples missing some variables: a GaussianImputer fitted to the data fills them afresh in every realization, so the imputation uncertainty reaches the realizations. Given an imputer, its components, spatial and neighbors are used, so with spatial the gaps also follow nearby samples. The transform is fitted to the complete samples. Samples missing every variable are dropped.

 None

from_parquet staticmethod 

from_parquet(path)

Reads to_parquet output; raises InvalidInput on another class's file or a newer format.

simulate method descriptor 

simulate(
    targets,
    *,
    n=100,
    seed=0,
    cutoffs=Ellipsis,
    quantiles=Ellipsis,
    keep=None,
    anisotropy=None,
    blocks=None,
)

Summaries of n realizations of every variable.

Parameters are those of SGS.simulate; anisotropy needs SGS for every factor. Factor j of realization k is simulated with a seed mixed from seed, k and j, so no two factors share random numbers.

Returns:

Type Description
list of SimulationSummary

One per variable, in column order.

to_parquet method descriptor 

to_parquet(path)

Writes arrays as Parquet columns and parameters as JSON in the file metadata; from_parquet reads it back.

used in