fips.covariance.CorrelationModel.with_std#

CorrelationModel.with_std(std)[source]#

Return the covariance of this correlation with standard deviations std.

Overloads:

Entry (i, j) is std[i] * C[i, j] * std[j]. This is the one way to make a covariance from a model.

Parameters:

std (float | Series | Vector) – The standard deviation of each label, not negative. A number is the same for every label. A Series is indexed by the labels of the one block the covariance is built on, without a block level (vector["flux"]’s labels, space.select("flux")), and is lined up with them by label when the model is built. A Vector brings its own space. A plain array is refused: it has no labels, so nothing would say which value is whose.

Returns:

For a Vector, the Covariance on its space. For a number or a Series, a CovarianceModel, which build() puts on a space.

Return type:

Covariance or CovarianceModel

Raises:

  • TypeError – If std is an array, a list or anything else that is not a number, a Series or a Vector.

  • ValueError – If a standard deviation is negative or not finite, or the Series’ index has a block level, an unnamed level or a label twice.

See also

fips.covariance.CovarianceModel.build

Put the model on a space.

Examples

>>> import pandas as pd
>>> from fips import Exponential, Space, Vector
>>> x = pd.Index([0.0, 1.0, 2.0], name="x")
>>> space = Space({"flux": x})
>>> S = Exponential("x", 1.0).with_std(2.0).build(space)
>>> S.dense().round(3)
array([[4.   , 1.472, 0.541],
       [1.472, 4.   , 1.472],
       [0.541, 1.472, 4.   ]])

A Series is read by label, whatever its order:

>>> std = pd.Series([2.0, 1.0, 1.0], index=x[::-1])
>>> Exponential("x", 1.0).with_std(std).build(space).variances.values
array([1., 1., 4.])

A Vector is a covariance at once:

>>> std = Vector([1.0, 1.0, 2.0], space)
>>> Exponential("x", 1.0).with_std(std).variances.values
array([1., 1., 4.])