tfp.experimental.timeseries.horizoned_msis

Return the mean scaled interval score over all horizons.

tfp.experimental.timeseries.horizoned_msis(
    y, yhat_lower, yhat_upper, y_obs, m, a=0.025
)

Args
y	An array containing the true values.
yhat_lower	An array containing the a% quantile of the predicted distribution.
yhat_upper	An array containing the (1-a)% quantile of the predicted distribution.
y_obs	An array containing the training values.
m	The season length.
a	A scalar in [0, 1] specifying the quantile window to evaluate.

Returns
A list a, with a[i] containing the MSIS over y[0] .. y[i].