kedm.ccm

kedm.ccm(lib: numpy.ndarray[numpy.float32], target: numpy.ndarray[numpy.float32], *, lib_sizes: List[int] = [], sample: int = 1, E: int = 1, tau: int = 1, Tp: int = 0, seed: int = 0, accuracy: float = 1.0) → List[float]

Estimate the strength of causal interaction between two time series using Convergent Cross Mapping (CCM).

Parameters:

  • lib – Library time series

  • target – Target time series

  • lib_sizes – List of library sizes

  • sample – Number of random samples

  • E – Embedding dimension

  • tau – Time delay

  • Tp – Prediction interval

  • seed – Random seed (randomly initialized if 0)

  • accuracy – Approximation accuracy

Returns:

List of Pearson’s correlation coefficient for each library size

Note

If accuracy < 1.0, approximate nearest neighbor search is used to speed up execution with a slightly reduced accuracy. For example, 99.9% of the true neighbors is expected to be used if accuracy is set to 0.999.