pflm.fpca.utils.get_fpca_in_score#

get_fpca_in_score(flatten_func_data: FlattenFunctionalData, mu: ndarray, num_pcs: int, fpca_lambda: ndarray, fpca_phi: ndarray, sigma2: float, if_shrinkage: bool = False) → tuple[ndarray, list[ndarray], ndarray, list[ndarray]][source][source]#

Compute Numerical integration FPCA scores and fitted curves.

Parameters:

flatten_func_dataFlattenFunctionalData: Flattened data containing fields y, t, tid, unique_sid, sid_cnt.
munp.ndarray of shape (nt,): Mean on the observation grid.
num_pcsint: Number of principal components to use (<= len(fpca_lambda)).
fpca_lambdanp.ndarray of shape (k,): FPCA eigenvalues.
fpca_phinp.ndarray of shape (nt, k): Basis on the observation grid (columns are components).
sigma2float: Measurement noise variance used in shrinkage (if enabled).
if_shrinkagebool, default=False: Whether to apply shrinkage to the IN scores.

Returns:

xinp.ndarray of shape (n_samples, num_pcs): IN scores by subject.
xi_varList[np.ndarray]: Per-subject score covariance matrices or variance summaries.
fitted_y_matnp.ndarray of shape (nt, n_samples): Fitted values on the observation grid.
fitted_yList[np.ndarray]: Fitted values at observed time points per subject.

Raises:

ValueError: If num_pcs exceeds available eigenvalues, fpca_phi has incompatible shape, or if_shrinkage is not a boolean.