AGO 2011-15

[QLLZ]

Data pull: HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8 config

[QLLZ]

Model run: HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8

[QLLZ]

Country data report

[eclee25]

The model chains look like they have converged, except the variation in rho, w, and sd_w posterior samples seems to be increasing in later parts of the trace plot. In addition, while the mean annual incident case maps look okay and match the GAM input quite well, there does not seem to be good concordance between the mean annual incidence rate estimates and the corresponding GAM input. Subnational observations seem sparse, so I wonder if the model had difficulty allocating cases spatially.

Opinion: Model output files and data comparison report need further investigation.

[javierps]

I agree that the differences may be due to the low amount of sub-national data/ large parts of the country not being covered. This is probably also the reason behind slight convergence issues with rho and sd_w. I don't think we can do much about it.

What would you suggest looking for in the output files? @eclee25

Given that validation and convergence are OK I be ok with validating the report.

[eclee25]

Moved to Approve column after second review

[QLLZ]

Config

Data pull & Model run:

HASH: 211913e0eb7d570ed62c48bc1ae9ecd6671a6cbc

[QLLZ]

country data report

[QLLZ]

Model cases look good. There's slightly convergency issue, probably due to few subnational observations which is consistent with previous reports. Population look good compared to wpp.

[eclee25]

Convergence issues with w, sd_w, and rho but other diagnostics look okay. Issues could be related to having few subnational observations.

Opinion: Reject and rerun with do_sd_w_mixture: no

[QLLZ]

Updated config

[QLLZ]

Data pull & model run:

HASH: 211913e0eb7d570ed62c48bc1ae9ecd6671a6cbc

[QLLZ]

Updated country data report

[eclee25]

AGO traceplots are a bit improved without the mixture sd_w prior but I do wonder if a lower rho prior might help

[javierps]

Agreed. I think what could help is to put a stronger prior on std_dev_w, but not sure this will solve everything as there are large parts of the country without any data.

[eclee25]

ID-57 run is approved. For future runs, we need to have do_sd_w_mixture: no Approve ID-57

[QLLZ]

Data pull, model run:

HASH: 05f598b95b5d42ac0049963f8a0b7e04d6f1dbd7

[javierps]

Issues with convergences in rho, std_w and ws OK. Rhats obs OK. Rhats output slight issues at adm 2 with all values < 1.1.

Suggestion: Accept.

[eclee25]

Approve