neurostatx.utils.factor

Classes

Name	Description
MethodTypes	Factor extraction methods for exploratory factor analysis.
RotationTypes	Factor rotation methods for exploratory factor analysis.

Functions

Name	Description
cfa	Fit a confirmatory factor analysis model with semopy.
efa	Fit an exploratory factor analysis with factor_analyzer.
horn_parallel_analysis	Estimate the number of factors and components with Horn's method.

MethodTypesclass#

View source

class MethodTypes(StrEnum, Enum)

Bases: StrEnum, Enum

Factor extraction methods for exploratory factor analysis.

minresattributeclass attributeinstance attribute#

View source

minres = 'minres'

mlattributeclass attributeinstance attribute#

View source

ml = 'ml'

principalattributeclass attributeinstance attribute#

View source

principal = 'principal'

RotationTypesclass#

View source

class RotationTypes(StrEnum, Enum)

Bases: StrEnum, Enum

Factor rotation methods for exploratory factor analysis.

equamaxattributeclass attributeinstance attribute#

View source

equamax = 'equamax'

oblimaxattributeclass attributeinstance attribute#

View source

oblimax = 'oblimax'

obliminattributeclass attributeinstance attribute#

View source

oblimin = 'oblimin'

promaxattributeclass attributeinstance attribute#

View source

promax = 'promax'

quartimaxattributeclass attributeinstance attribute#

View source

quartimax = 'quartimax'

quartiminattributeclass attributeinstance attribute#

View source

quartimin = 'quartimin'

varimaxattributeclass attributeinstance attribute#

View source

varimax = 'varimax'

cfafunction#

View source

def cfa(df, model)

Fit a confirmatory factor analysis model with semopy.

Parameters

Name	Type	Description
df	pd.DataFrame	Observed indicators included in the CFA.
model	str	semopy model specification.

Returns

fittedsemopy.Model

Fitted CFA model.

scorespd.DataFrame

Predicted factor scores.

statspd.DataFrame

Parameter estimates from inspect.

Examples

>>> import pandas as pd
>>> import numpy as np
>>> from neurostatx.utils.factor import cfa
>>> rng = np.random.RandomState(0)
>>> f = rng.randn(40)
>>> df = pd.DataFrame({"x1": f + 0.1 * rng.randn(40),
...                    "x2": f + 0.1 * rng.randn(40),
...                    "x3": f + 0.1 * rng.randn(40)})
>>> fitted, scores, stats = cfa(df, "F =~ x1 + x2 + x3")
>>> "F" in scores.columns
True

efafunction#

View source

def efa(df, method, rotation, nfactors=1)

Fit an exploratory factor analysis with factor_analyzer.

Parameters

Name	Type	Default	Description
df	pd.DataFrame	required	Variables to include in the EFA.
method	str	required	Extraction method: "minres", "ml", or "principal".
rotation	str	required	Rotation applied to the loadings, or None.
nfactors	int	1	Number of factors to extract. Defaults to 1.

Returns

modelFactorAnalyzer

Fitted factor analyzer.

evnp.array

Original eigenvalues.

vnp.array

Common-factor eigenvalues.

scoresnp.array

Factor scores.

loadingsnp.array

Factor loadings.

communalitiesnp.array

Communalities.

Examples

>>> import pandas as pd
>>> import numpy as np
>>> from neurostatx.utils.factor import efa
>>> rng = np.random.RandomState(0)
>>> df = pd.DataFrame(rng.rand(40, 4), columns=list("abcd"))
>>> model, ev, v, scores, loadings, communalities = efa(
...     df, method="minres", rotation="varimax", nfactors=1
... )
>>> scores.shape[0]
40

horn_parallel_analysisfunction#

View source

def horn_parallel_analysis(x, output_folder, method='minres', rotation=None, nfactors=1, niter=20)

Estimate the number of factors and components with Horn’s method.

Compares observed eigenvalues to those from random data, following psych’s fa.parallel.

Parameters

Name	Type	Default	Description
x	np.array	required	Variables to include in the analysis.
output_folder	str	required	Directory where horns_parallel_screeplot.png is written.
method	str	'minres'	Extraction method: "minres", "ml", or "principal". Defaults to "minres".
rotation	str	None	Rotation applied when fitting. Defaults to None.
nfactors	int	1	Number of factors extracted while computing eigenvalues. Defaults to 1.
niter	int	20	Number of random-data iterations. Defaults to 20.

Returns

suggfactorsint

Suggested number of factors.

suggcomponentsint

Suggested number of components.

Examples

>>> import numpy as np
>>> from neurostatx.utils.factor import horn_parallel_analysis
>>> rng = np.random.RandomState(0)
>>> x = rng.rand(40, 5)
>>> suggfactors, suggcomponents = horn_parallel_analysis(
...     x, output_folder=".", niter=2
... )