spcal.calc

Misc and helper calculation functions.

spcal.calc.expand_mask(mask: ndarray, size: int) → ndarray

Grows mask values in mask by size on either side.

Used to expand the regions where peaks are detected, to exclude nearby data. This is equivalent to scipy.ndimage.binary_dilation except faster.

Parameters:

• mask – a 1d or 2d-array of boolean type of shape (samples, features)

• size – size to expand on both sides

Returns:

mask dilated by size, same shape as mask

spcal.calc.interpolate_3d(x: ndarray | float, y: ndarray | float, z: ndarray | float, xs: ndarray, ys: ndarray, zs: ndarray, data: ndarray) → ndarray | float

Cubic interpolation of (x, y, z) for data.

Parameters:

• x – x position of values to interpolate

• y – y position of values to interpolate

• z – z position of values to interpolate

• xs – x values of data

• ys – y values of data

• zs – z values of data

• data – known values, shape (xs, ys, zs)

Returns:

interpolated values, shape (x)

spcal.calc.is_integer_or_near(x: ndarray | float, max_deviation: float = 0.001) → ndarray | float

Test if float data is ‘near’ integer. Near integers values are those less than max_deviation from a whole number.

Parameters:

• x – float array

• max_deviation – max distance from whole number

Returns:

array of bool

spcal.calc.mode(x: ndarray, bins: int | ndarray | str = 'auto') → float

Calculates the mode of ‘x’.

Parameters:

x – array

Returns:

mode

spcal.calc.otsu(x: ndarray, remove_nan: bool = False, nbins: str | int = 'fd') → float

Calculates the otsu threshold.

The Otsu threshold minimises intra-class variance for a two class system. If remove_nan then all nans are removed before computation.

Parameters:

• x – array

• remove_nan – remove nan values

• nbins – number of bins to use

See also

skimage.filters.threshold_otsu()

spcal.calc.pca(x: ndarray, trim_to_components: int = 2) → tuple[ndarray, ndarray, ndarray]

Perform a PCA on ‘x’, standard scales data.

Parameters:

• x – input array of shape (samples, features)

• trim_to_components – trim to this many dims

Returns:

pca data points component vectors explained variance per dim

spcal.calc.search_sorted_closest(x: ndarray, v: ndarray, check_max_diff: float | None = None)

Get the idx of the closest values in x for v.

If check_max_diff is a value, the maximum distance must be lower.

Parameters:

• x – sorted array

• y – values to find closest idx of in x

• check_max_diff – if not None, check maximum diff is less than this

Returns:

idx of closest x values for v

Raises:

ValueError if check_max_diff is not None and max diff greater. –

spcal.calc.weighted_linreg(x: ndarray, y: ndarray, w: ndarray | None = None) → tuple[float, float, float, float]

Weighted linear regression.

Uses polyfit with sqrt(weights) for intercept and gradient.

Parameters:

• x – 1d-array

• y – array, same size as x

• w – weights, same size as x

Returns:

gradient intercept r² error, S(y,x) the (unweighted) residual standard deviation

See also

pewlib.calibration.weighted_rsq()

spcal.calc.weighted_rsq(x: ndarray, y: ndarray, w: ndarray | None = None) → float

Calculate r² for weighted linear regression.

Parameters:

• x – 1d-array

• y – array, same size as x

• w – weights, same size as x

spcal.calc.weights_from_weighting(x: ndarray, weighting: str, safe: bool = True) → ndarray

Get weighting for x.

Conveience function for calculating simple weightings. If safe then any zeros in x are replace with the minimum non-zero value.

Parameters:

• x – 1d-array

• weighting – weighting string {‘equal’, ‘x’, ‘1/x’, ‘1/(x^2)’}

• safe – replace zeros with minimum

Returns:

weights, same size as x