utilities.estimate_region_weights()
Estimate regional weights from the growth data alone.
Usage
utilities.estimate_region_weights(
y_uk,
y_annual,
shrinkage=0.2,
)National quarterly growth is aggregated to annual growth with the same weights (OMEGA) that link quarterly and annual regional growth, and regressed on the regions’ annual growth with weights constrained to be non-negative and sum to one. The estimate is shrunk towards equal weights because regional growth rates move together and there are few annual observations, so the data pin the weights down only loosely. Known shares (e.g. of output) are a better source when available.
Parameters
y_uk: npt.NDArray[np.float64]-
National quarterly growth, shape (T,).
y_annual: npt.NDArray[np.float64]-
Regional annual growth, shape (T, R), NaN where unobserved.
shrinkage: float = 0.2-
Strength of the pull towards equal weights, relative to the scale of the data. Zero means none.
Returns
npt.NDArray[np.float64]-
npt.NDArray[np.float64]: Estimated shares, shape (R,), summing to one. Equal weights if no quarter has every region observed.