Maciej Beręsewicz

- [ ] calibration for quantiles as described in https://arxiv.org/abs/2403.09726 - [ ] propensity score: empirical likelihood - [ ] propensity score: maximum entropy - [ ] propensity score: non-parametric...

For `gaussian` I got ```r > nonprob( + outcome = single_shift~region + private + nace + size, + pop_totals = pop_totals, + method_outcome = "glm", + data = admin) Caution:...

Add support for other distances and high performance via the the `rnndescent` package

Add options: - [ ] `KernSmooth` -- continuous, low dim - [ ] `np` -- mixed data

Prediction approach

1

What is missing: + provide population data (either in a form of population data or a data.frame with frequencies) + list of estimators: model-calibration, mass imputation

+ Theory in [ILR_and_PILR.pdf](https://github.com/user-attachments/files/21899970/ILR_and_PILR.pdf) Initial implementation here: 1. implementation of the IRL estimator ```r est_ilr