calpit.utils

Functions

normalize(→ numpy.ndarray)	Normalizes conditional density estimates to be non-negative and integrate to one.
_normalize(density, y_grid[, tol, max_iter])
trapz_grid(→ numpy.ndarray)	Does trapezoid integration between the same limits as the grid.
trapz_grid_torch(→ torch.tensor)	Same as trapz_grid but implemented in Pytorch
plot_pit(pit_values, ci_level[, n_bins, y_true, ax])	Plots the PIT/HPD histogram and calculates the confidence interval for the bin values,

Module Contents

normalize(cde_estimates: numpy.ndarray, y_grid: numpy.ndarray, tol: float = 1e-06, max_iter: int = 200) → numpy.ndarray

Normalizes conditional density estimates to be non-negative and integrate to one.

Parameters:

• cde_estimates (numpy.ndarray) – A numpy array or matrix of conditional density estimates.

• x_grid (numpy.ndarray) – The array of grid points.

• tol (float) – The tolerance to accept for abs(area - 1).

• max_iter (int) – The maximal number of search iterations.

Returns:

The normalized conditional density estimates.

Return type:

numpy.ndarray

_normalize(density, y_grid, tol=1e-06, max_iter=500)

trapz_grid(y: numpy.ndarray, x: numpy.ndarray) → numpy.ndarray

Does trapezoid integration between the same limits as the grid.

Parameters:

• y (np.ndarray) – The array of values to integrate.

• x (np.ndarray) – The array of grid points.

Returns:

The integrated values.

Return type:

np.ndarray

trapz_grid_torch(y: torch.tensor, x: torch.tensor) → torch.tensor

Same as trapz_grid but implemented in Pytorch

plot_pit(pit_values, ci_level, n_bins=30, y_true=None, ax=None, **fig_kw)

Plots the PIT/HPD histogram and calculates the confidence interval for the bin values, were the PIT/HPD values follow an uniform distribution

@param values: a numpy array with PIT/HPD values @param ci_level: a float between 0 and 1 indicating the size of the confidence level @param x_label: a string, populates the x_label of the plot @param n_bins: an integer, the number of bins in the histogram @param figsize: a tuple, the plot size (width, height) @param ylim: a list of two elements, including the lower and upper limit for the y axis @returns The matplotlib figure object with the histogram of the PIT/HPD values and the CI for the uniform distribution