Open eclee25 opened 1 year ago
Model run:
HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8
Some instability sd_w and w (trend in samples over time) and some rhat issues. Maps, fits, GAM input comparison look ok. 2011 is the only year with non-zero cases and case counts are generally low.
Opinion: Leaning towards Approve, need second opinion
The issue here is the lack of sub-national level data, so we don't expect the w estimates to be meaningful.
I would suggest approving as the posterior probs of the risk categories at sub-national level do seem OK.
Opinion: Temp Approve and rerun with spatial_effect: no
and higher sfrac_thresh_border due to highish rates in N and W
This run was not moved to no-spatial effect column. Rerun with no-w.
Results look good.
Suggestion: accept.
Almost no subnational data. no-W model run looks good. Approve
As there is some subnational data in the year with cases, and the sd_w was trending upward in the standard model, let's try with the no-mixture sd_w prior
Data pull, model run:
HASH: 05f598b95b5d42ac0049963f8a0b7e04d6f1dbd7
Issues with rhat, small drift in std_w, ws also seem to have slight issues. Rhats OK except for one datapoint at 1.06, output rhats mostly OK with max values < 1.1. Given the amount of data not sure we can do much better.
Suggestion: Accept.
Investigate 2011 censored observations -- why are they underestimated?
Very minimal subnational data and national observation fits are ok. Discussed as a group. Approve
Data pull: HASH:
6772ea9da4f43ddc3da11c494a573af2ca3dd3c8
config