search

krennpa / SAEforest

https://krennpa.github.io/SAEforest/
GNU General Public License v2.0
3 stars 1 forks source link

Estimate jackknife standard errors #3

Open IsaMariaSteiner opened 1 year ago

IsaMariaSteiner commented 1 year ago

This is a wonderful package!

However, it does not seem possible to estimate standard errors for the predicted values.

Technically, this is possible in ranger: https://github.com/imbs-hl/ranger/issues/136

Would it be possible to implement it in the predict function of SAEforest as well?

krennpa commented 1 year ago

Hi there and thank you for enjoying the package, The package has a focus on the estimation of regional indicators and the MERFranger function was thought to allow experimenting with individual point estimates. The difficulty in producing reliable standard errors for the individual predicted point estimates for MERFs is to additionally capture the variability of the random effects part of the model. We were able to do this by a block-bootstrap method for area-level means: https://doi.org/10.1111/rssc.12600. However, to the best of my knowledge there currently exists no reliable method to assess the variability of individual predicted values.