MMEFM 0.1.1
- Rename the simulation time-length argument from
T to
TT and the internal local common-factor variable to comply
with CRAN naming guidance. Statistical calculations and returned object
schemas are unchanged.
- Simulate grouped matrix-valued time series with heterogeneous
dimensions, Gaussian or raw Student-t innovations, and optional
model-component storage.
- Generate piecewise ARFIMA series with stable AR(p) dynamics and
shared or series-specific fractional memory.
- Estimate global and local main effects and common components at
supplied or automatically selected ranks, with Van Loan alignment and
refitting by default and optional Procrustes alignment and ALS
refitting.
- Select the eight loading rank classes using unperturbed eigenvalue
ratios, indexwise cross-group maxima for global spectra, and full
consecutive-ratio search ranges. Check combined ranks against the
IC1-centered spatial dimensions without rank repair.
- Return structured model fits with print and summary methods,
fitted-component reconstruction, and final residuals.
- Detect global common factors and screen participating groups using a
shared circular-shift bootstrap cutoff, with optional PSOCK execution
and matching serial/parallel random inputs.