MOZ 2011-15

[QLLZ]

Data pull: HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8 config

[QLLZ]

Model run: HASH: 6772ea9da4f43ddc3da11c494a573af2ca3dd3c8

[QLLZ]

Country data report

[eclee25]

sd_w and rhat have some convergence issues, lots of Rhat issues for genquant. Rate maps look oddly uniform within a single year, but observation fits look okay.

Opinion: Convergence issues and rate maps need further investigation.

[javierps]

The issue here I think is that there is very little data at the subnational level, and the country is quite big. Note the very small range of w values.

I suspect that all sub-national data is low values compatible with a uniform incidence rate in those years. Would suggest making admin-level plots for the validation figures.

[javierps]

Suggestion: no-W run given the low amount of sub-national data with respect to the size of the country.

[eclee25]

Rerun on dev (with spatial_effect: yes) now that the DAGAR has been fixed.

[QLLZ]

Data run

HASH f56d5a942b106ad7750b9be31c0456cefa5b6dc7

[QLLZ]

model run

HASH f56d5a942b106ad7750b9be31c0456cefa5b6dc7

[QLLZ]

Updated configs

[QLLZ]

Data pull and model run: HASH: fb20b87bab8e4c41fe520de2dc76de0050ece361

[QLLZ]

Country data report

[eclee25]

Continued convergence issues on the sd_w and w traceplots, which may be related to the mixture prior on sd_w. We will try to relaunch this on dev_nomixture_sd_w_prior branch with the config setting do_sd_w_mixture: no

[eclee25]

When do_sd_w_mixture: no the sd_w trace plots looked better although not perfect. This setting can be an exception to a standard run -- designed for countries with only limited subnational data.

[eclee25]

Moving this back to standard run because we are changing the mixture prior on sd_w to have a stronger weight on lower values

[QLLZ]

Config

Data pull & Model run:

HASH: 211913e0eb7d570ed62c48bc1ae9ecd6671a6cbc

[QLLZ]

country data report

[QLLZ]

Bad convergency (few subnational data). Modeled cases look good.

Suggestion: maybe try other settings?

[eclee25]

Convergence issues with w and sd_w even after mixture prior weight was shifted lower. Opinion: Reject. Rerun with do_sd_w_mixture: no

[QLLZ]

Updated config

[QLLZ]

Data pull & model run:

HASH: 211913e0eb7d570ed62c48bc1ae9ecd6671a6cbc

[QLLZ]

Updated country data report

[eclee25]

MOZ traceplots are much improved even if the non-mixture sd_w prior hasn’t completely resolved everything.

[javierps]

Agreed, the Rhats are higher not super large, and I don't think we will be able to do much better here unless we put a stronger prior on std_dev_w.

[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]

Rho convergence not perfect, std_w OK, ws seem to also have some differences. Rhats all OK.

Suggestion: Accept.

[eclee25]

Approve