Compute Eigenvalue Spectrum Fidelity Metrics

Compares subject-specific eigenvalues from HATSA (`Lambda_original_list`) with true eigenvalues (e.g., from a toy data generator).

Usage

compute_eigenvalue_fidelity(
  hatsa_object,
  true_eigenvalues_list,
  k_to_compare = NULL
)

Arguments

hatsa_object

A fitted `hatsa_projector` object, which contains `Lambda_original_list` (list of numeric vectors of eigenvalues).

true_eigenvalues_list

A list of numeric vectors, where each vector contains the true eigenvalues for the corresponding subject. If a single vector is provided, it is assumed to be the common true eigenvalues for all subjects.

k_to_compare

Integer or NULL. The number of top eigenvalues to compare. If NULL (default), all available eigenvalues are compared (up to the length of the shorter of the estimated or true eigenvalue vectors for that subject).

Value

A list of lists, one for each subject. Each inner list contains:

correlation

Pearson correlation between estimated and true eigenvalues.

mse

Mean Squared Error between estimated and true eigenvalues.

num_compared

The number of eigenvalue pairs actually compared.

Returns NULL for a subject if inputs are invalid for that subject.

Examples

# See core-hatsa-toy-example.Rmd for usage.