neurostatx.cli.LogisticRegression

Attributes

Name	Description
app	Cyclopts application for the LogisticRegression command-line tool.

Functions

Name	Description
LogisticRegression	Fit logistic regression of graph edges against node attributes.

appattributemodule attribute#

View source

app = App(default_parameter=Parameter(negative=()))

Cyclopts application for the LogisticRegression command-line tool.

LogisticRegressionfunction#

View source

def LogisticRegression(in_graph: Annotated[str, Parameter(show_default=False, group='Essential Files Options')], out_folder: Annotated[str, Parameter(show_default=False, group='Essential Files Options')], attributes: Annotated[List[str], Parameter(group='Model Parameters', show_default=False)], covariates: Annotated[List[str], Parameter(group='Model Parameters', show_default=True)] = None, weight: Annotated[str, Parameter(group='Model Parameters', show_default=True)] = 'membership', splits: Annotated[int, Parameter(group='Model Parameters', show_default=True)] = 10, test_size: Annotated[float, Parameter(group='Model Parameters', show_default=True)] = 0.2, cs: Annotated[int, Parameter(group='Model Parameters', show_default=True)] = 10, max_iter: Annotated[int, Parameter(group='Model Parameters', show_default=True)] = 1000, penalty: Annotated[Penalty, Parameter(group='Model Parameters', show_default=True, show_choices=True)] = Penalty.l2, solver: Annotated[Solver, Parameter(group='Model Parameters', show_default=True, show_choices=True)] = Solver.lbfgs, permutations: Annotated[int, Parameter(group='Model Parameters', show_default=True)] = 1000, scoring: Annotated[ScoringMethod, Parameter(group='Model Parameters', show_default=True)] = ScoringMethod.r2, processes: Annotated[int, Parameter(group='Model Parameters', show_default=True)] = 1, plot_distributions: Annotated[bool, Parameter(show_default=True, group='Visualization Options')] = False, verbose: Annotated[bool, Parameter(-v, --verbose, group='Optional parameters')] = False, save_parameters: Annotated[bool, Parameter(-s, --save_parameters, group='Optional parameters')] = False, overwrite: Annotated[bool, Parameter(-f, --overwrite, group='Optional parameters')] = False)

Fit logistic regression of graph edges against node attributes.

LogisticRegression uses edge weights as predictors and node attributes as the response. It cross-validates on a training set, tests on held-out data, then runs permutation testing for statistical significance. It writes coefficients, statistics, and plots of attribute and edge-weight distributions and of the coefficients.

Notes

Preprocessing

Data are scaled to unit variance and zero mean. Edge weights are log-transformed; for now this assumes the weights are membership values from a fuzzy clustering analysis.

Node attributes

The script takes a single .gml graph file, fetches node attributes from it, and fits the model on those attributes and the edge weights. At least one attribute is required. To attach attributes to nodes, see AddNodesAttributes.

Scoring

Permutation testing assesses whether the model is statistically significant. The p-value uses area under the curve (AUC) by default; other scores are listed in the scikit-learn scoring documentation [1]. The single-tailed p-value is:

p-value = ∑(score_perm >= score) / (nb_permutations)

Coefficient significance

Coefficient p-values are also computed from the permutation distribution by comparing model coefficients to permuted coefficients. The two-tailed p-value is:

p-value = ∑(abs(coef_perm) >= abs(coef)) / (nb_permutations)

References

[1] scikit-learn scoring parameter

Parameters

Name	Type	Default	Description
in_graph	str	required	Graph file containing the data for the model.
out_folder	str	required	Output folder.
attributes	List[str]	required	Attribute names to include in the model. Must be present in the graph file. At least one attribute is required.
covariates	List[str]	None	Covariates to include in the model. Must be present in the graph file. Defaults to None.
weight	str	'membership'	Edge weight to use for the model. Defaults to membership.
splits	int	10	Number of splits to use for the cross-validation. Defaults to 10.
test_size	float	0.2	Size of the testing set. Must be between 0 and 1. Defaults to 0.2.
cs	int	10	Inverse of regularization strength. Smaller values specify stronger regularization. Defaults to 10.
max_iter	int	1000	Maximum number of iterations for the solver. Defaults to 1000.
penalty	Penalty	Penalty.l2	Regularization penalty to use for the LogisticRegression model. Defaults to l2.
solver	Solver	Solver.lbfgs	Solver to use for the LogisticRegression model. Defaults to lbfgs.
permutations	int	1000	Number of permutations to use for the permutation testing. Defaults to 1000.
scoring	ScoringMethod	ScoringMethod.r2	Scoring method to use for the permutation testing. Defaults to r2.
processes	int	1	Number of processes to use for the cross-validation. Defaults to 1.
plot_distributions	bool	False	If true, will plot the distributions of the edges’ weights. Defaults to False.
verbose	bool	False	If true, produce verbose output. Defaults to False.
save_parameters	bool	False	If true, will save input parameters to .txt file. Defaults to False.
overwrite	bool	False	If true, force overwriting of existing output files. Defaults to False.

Examples

LogisticRegression --in-graph graph.gexf --out-folder output_folder -v
-s