MetaWards wave 03 pre-lockdown emulation test

Emulation tests for MetaWards

Model description (parameters etc. ) are here

Based on Code from James Salter.

Code for this notebook is here

Ensemble design

Cumulative hospital deaths at the start of lockdown against each parameter

Cumulative community deaths at the start of lockdown against each parameter

Emulating cumulative deaths in hostpital (cumHD) in the community (cumCD) at the beginning of lockdown

The DiceKriging Emulator fit for log(deaths) looks good. Leave-one-out isn’t a great measure, but it’s a start.

Week 12 cumulative Hospital deaths Leave-one-out plots

Week 12 cumulative community deaths Leave-one-out plots

Normalise input space and re-fit to produce some sensitivity analyses

one-at-a-time sensitivity for cumulative hospital deaths

one-at-a-time sensitivity for cumulative community deaths

FAST99 sensitivity analysis of Saltelli et al (1999)

Call:
fast99(model = NULL, factors = colnames(X_norm), n = 3000, q = "qunif",     q.arg = list(min = 0, max = 1))

Model runs: 42000 

Estimations of the indices:
          first order total order
R0       0.0357894898 0.052647577
TE       0.0094237203 0.015631583
TP       0.0111223776 0.016713101
TI1      0.0042119359 0.007605358
TI2      0.0005845405 0.002777098
nuA      0.0017362075 0.004688977
ns       0.0040062589 0.006783809
alphaTH  0.0108279236 0.012842142
etaTH    0.0167128822 0.019462817
alphaEP  0.0702844368 0.072592994
alphaI1D 0.0018994493 0.004351074
alphaHD  0.6488963359 0.676379829
alphaI1H 0.0509419175 0.053030910
eta      0.0994782982 0.101617848

Call:
fast99(model = NULL, factors = colnames(X_norm), n = 3000, q = "qunif",     q.arg = list(min = 0, max = 1))

Model runs: 42000 

Estimations of the indices:
          first order  total order
R0       3.707727e-02 0.0488105154
TE       1.159102e-02 0.0140520108
TP       1.114574e-02 0.0133480137
TI1      4.404289e-03 0.0053919290
TI2      7.830393e-05 0.0008514642
nuA      2.755183e-03 0.0039733774
ns       2.318849e-03 0.0036564024
alphaTH  1.607502e-05 0.0005314890
etaTH    1.100540e-05 0.0004947353
alphaEP  5.548157e-02 0.0565196581
alphaI1D 8.110668e-01 0.8287016527
alphaHD  8.497626e-05 0.0011878537
alphaI1H 4.147824e-04 0.0033935812
eta      4.080266e-02 0.0419101180