SOM 2016-20

[eclee25]

No description provided.

[QLLZ]

Data pull:
HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8
config

[QLLZ]

Model run:
HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8

[QLLZ]

country data report failed: log file

[javierps]

This seems to be due to the gam input object in the grid-level variance plot.
@QLLZ could you check the values that are passed to the plotting function?

[QLLZ]

@javierps Yeah, I changed the grid-level_variance plot and a new issue came up. it's about the w value: the length of w value is different in sf_grid and model.rand. After checking the log file of the model run, i found that only 1 chain successfully finished, chain 1, 3-4 actually faild.

[javierps]

Thanks @QLLZ .

Stan model failure is possibly due to the same initialization issue in COD 2016.

[eclee25]

Rerun on dev with drop_multiyear_adm0: yes config option

[QLLZ]

Updated config

[QLLZ]

Data pull:
HASH 02faaed1a21116cebc29533c68ec0d99fe78d08f

[QLLZ]

Model run:
HASH 02faaed1a21116cebc29533c68ec0d99fe78d08f

[QLLZ]

Country data report

[eclee25]

Questioning whether model was run on correct branch with correct taxdat installation as there are still multi-year admin 0 obs in the stan input when these should have been dropped.

Opinion: rerun after pull / reinstallation

[QLLZ]

rerun on dev:
HASH; a4d49235ab2737ed5eaafe23264a831ab9cceef7

[QLLZ]

Model run on dev:
HASH: a4d49235ab2737ed5eaafe23264a831ab9cceef7

[QLLZ]

Similar issue as before, only 1 chain succeed.

log file

[QLLZ]

Data pull:
HASH: 23e2dc77a189bac73d6b4359bc83d99dfb99cf2f

[QLLZ]

Model run:
HASH: 23e2dc77a189bac73d6b4359bc83d99dfb99cf2f

[QLLZ]

3 chains failed due to initial values

[QLLZ]

Updated config

[QLLZ]

Data pull:
HASH: d6950c7afd2b12c35abe6d5321ddfd31a75b2cb0

[QLLZ]

Model run on dev_drop_adm0_censored
HASH: d6950c7afd2b12c35abe6d5321ddfd31a75b2cb0

[QLLZ]

Mode run failed: log file

Data pull log

[eclee25]

After drop_censored_adm0 and drop_multiyear_adm0 were turned on (with drop_censored_adm0_thresh: 1), the model was able to initialize and the outputs converged and look good.

Opinion: Exception run with drop_censored_adm0: yes and drop_censored_adm0_thresh: 1

[eclee25]

Standard Run but with drop_censored_adm0_thresh: 1

[QLLZ]

Config

Data pull & Model run:

HASH: 211913e0eb7d570ed62c48bc1ae9ecd6671a6cbc

[QLLZ]

failed model run log file

[QLLZ]

Rerun on dev_u_combs_fix

HASH: e56580fa5ddab00293e31ff90139351f80ae3f6c

[QLLZ]

country data report

[QLLZ]

Same high incidence rate map as the other time period. Same pop estimates issue.

Modeled cases look good. Convergency looks good.

[eclee25]

All diags look good but pop estimates are way too low. Will adjust in post processing

Approve

[eclee25]

Standard settings with drop_censored_adm0_thresh: 1.0

[eclee25]

With the new dummy variable censoring implementation, this model now works with the standard settings. No need to use a non-standard drop_censored_adm0_thresh parameter.

[javierps]

Sep 2023 Production run: all OK.

Suggestion: accept.

[eclee25]

Approve