adam_core.coordinates.variants module¶
- class adam_core.coordinates.variants.VariantCoordinatesTable(*args, **kwargs)[source]¶
-
A protocol for a generic table of variant coordinates.
- property index: Int64Array¶
- property sample: T¶
- property weight: DoubleArray¶
- property weight_cov: DoubleArray¶
- adam_core.coordinates.variants.create_coordinate_variants(coordinates: CoordinateType, method: Literal['auto', 'sigma-point', 'monte-carlo'] = 'auto', num_samples: int = 10000, alpha: float = 1, beta: float = 0, kappa: float = 0, seed: int | None = None) VariantCoordinatesTable[CoordinateType][source]¶
Sample and create variants for the given coordinates by sampling the covariance matrices. There are three supported methods:
sigma-point: Sample the covariance matrix using sigma points. This is the fastest method, but can be inaccurate if the covariance matrix is not well behaved.
monte-carlo: Sample the covariance matrix using a monte carlo method. This is the slowest method, but is the most accurate.
auto: Automatically select the best method based on the covariance matrix. If the covariance matrix is well behaved then sigma-point sampling will be used. If the covariance matrix is not well behaved then monte-carlo sampling will be used.
When sampling with monte-carlo, 10k samples are drawn. Sigma-point sampling draws 13 samples for 6-dimensional coordinates.
Warning
This function does not yet handle sampling of covariances and coordinates with missing values.
- Parameters:
coordinates – The coordinates to sample.
method – The method to use for sampling the covariance matrix. If ‘auto’ is selected then the method will be automatically selected based on the covariance matrix. The default is ‘auto’.
num_samples (int, optional) – The number of samples to draw when sampling with monte-carlo.
alpha – Spread of the sigma points between 1e^-2 and 1.
beta – Prior knowledge of the distribution when generating sigma points usually set to 2 for a Gaussian.
kappa – Secondary scaling parameter when generating sigma points usually set to 0.
- Return type:
The variant coordinates.
- Raises:
ValueError: – If the covariance matrices are all undefined. If the input coordinates are not supported.