How bryan-cheong/rie-singular-value-cleaning-of-lagged-covariance-matrices works
Project type and purpose
• This is a standalone research/code‐analysis tool (not a web or mobile app) for “Singular Value Cleaning of Lagged Covariance Matrices” using the Rotationally Invariant Estimator (RIE).
File structure
• Top‐level scripts:
• RIE_visualisation.py – orchestrates data loading, RIE computation, and plotting.
• phi_AR.py – computes and validates φ‐distribution for AR(1) models.
• base_functions.py – core routines (matrix assembly, covariance, RIE algorithm).
• Data & output directory:
• data_and_figures/ – stores input data subsets (e.g. “AR(1)” and “RG”) and generated figures.
• Config & dependencies:
• requirements.txt – external libraries (NumPy, Matplotlib, SciPy, Pandas, Scikit‐Learn, Statsmodels).
• README.md – usage instructions and domain notes.
README for additional insights
• Usage patterns: both scripts are invoked from the command line.
• Configurable visualization parameters (figure size, save/display flags, models).
• Domain note: exponent discrepancy (–1/12 vs –1/6) documented for clarity.
Elements
– Data Store, Core Library, RIE Script, φ‐Script, External Libraries, Output Store
– Data → Core → Script → Output
– Scripts also invoke Core directly
– Blue for code modules, green for data stores, orange for external libs, purple for outputs
– “import”, “read data”, “compute RIE”, “plot/save figures”, “validate φ”