Balancing weights using (ne - oe).mean() vs base res?

[wkitlasten]

I notice the pst.adjust_weights() method ultimately uses pcf.base.rei to adjust the weights in the ies realm. In my case, those residuals are quite different than the mean residual (which I assume is appropriately calculated using (ne - oe).mean()?). Since the empirical jco is based on the ensemble mean, does it make sense to adjust the weights based on the ensemble based mean residual rather than the base realization? Or an I missing something (again)?

[jtwhite79]

There is a lot of subjectivity is all this weight stuff..and using the res file was easiest...you can always overwrite the values in pst.res with the ensemble mean values tho. We're you thinking of something else?

[jtwhite79]

Also FWIW the latest version of ies does this kind of weight adjustment internally and uses the mean residuals...

[wkitlasten]

Ha, cool. I see the ies_weight_ensemble() but it seems like that is something different? Any hints?

[jtwhite79]

https://github.com/usgs/pestpp/blob/develop/documentation/pestpp_users_manual.md#9210-internal-weight-adjustment

It's in the version 5.2 beta release candidate.

If you try it out, keen to hear how it goes - I've been using it quite a bit lately...

[wkitlasten]

Sounds cool. I took the "burdensome and potentially error prone" approach for now, but will give it a shot in a few days. The idea of having 'tags' in pestpp is cool.