Quantifying the importance of holiday-effect

Greetings!

Is there anyway to quantify holiday-importance using a scalar value for each?
I understand that the `model.predict()` output will contain the amount of the estimate `yhat` that is coming from that particular regressor (includes holidays). But these estimates are different for different points in time since the holiday-effect is a function of time _**h(t)**_.

Since under the hood holidays are essentially binary extra-regressors, can we pull their regression-coefficient or something like we do with `prophet.utilities.regression_coefficients()`?

_y(t) = g(t) + s(t) + **h(t)** + β*X(t) + ε_t_

#1487 suggested multiplying the X (matrix of Fourier component indicators) by the Betas (Fourier component Beta-coeffs):
`model.train_component_cols.T.dot(np.array(model.params['beta'])[0]) * model.y_scale`

Other rudimentary approaches I was thinking of:
1. Evaluate before and after adding each holiday (tedious and time-consuming).
2. Simple averaging of the holiday-columns in the `model.predict()` output (doesn't seem right to just average).

Am I interpreting this correctly?