report9

Main UK runs for Imperial Report 9 Tables 3,4,5 and A1

Results generated as indicated below should exactly match those in the *T8_NR10* output files in the GB_suppress and GB_mitigation folders (see CODECHECK certificate 2020-010 https://doi.org/10.5281/zenodo.3865491).

Results will be close to but not identical to those in the original report since: (a) the results produced here average over 10 stochastic realisations; (b) the population dataset has changed to Worldpop (open-source) rather than Landscan (closed); (c) the multi-threaded algorithm to create the simulation's household-to-place network has been modified to be single threaded and thus deterministic. Covidsim is now deterministic across platforms (linux, Mac and Windows) and across compilers (gcc, clang, intel and msvc) for a specified number of threads and fixed random number seeds.

Instructions common to both Suppression and Mitigation runs

Note: .ps1 files are Microsoft Powershell scripts and .sh are bash scripts.

• Get latest master of https://github.com/mrc-ide/covid-sim and compile with OpenMP enabled and floating point handling set to 'precise' (the cmake file in the repo does this).

• Run runonce.ps1 or runonce.sh in either the GB_mitigation or GB_suppress folder (no need to run this from both folders). This will create (in the population folder) a binary population file GB_pop2018.bin from the text population file GB_pop2018_nhs.txt and the household-place network file NetworkGB_8T.bin. It will also run a simulation for R0=2.4 with no interventions.

The GB_pop2018.bin and NetworkGB_8T.bin files produced by this run are required by both the suppression and mitigation runs. Note that other parameter files are not identical between the GB_suppress and GB_mitigation folders (due to differences in policy duration and triggering).

GB on/off trigger suppression policies in Imperial College NPI report (Report 9 Tables 4 and 5)

• Work in the GB_suppress folder

• Run batch.ps1 (on Windows) or batch.sh (on Linux) (script will require editing to substitute your favourite batch job scheduler to run on a compute cluster)

• Run summariseSup.r in R (changing cur_path at top of file first)

• Copy first 11 columns of meanT8_NR10\stats_contain.csv into the stats_contain sheet of a copy of stats_contain_meanT8_NR10.xlsx

• Refresh the pivot table in sheet Tables of that spreadsheet.

Original results are in stats_contain_orig.xlsx. Example results for linux and Windows are in files stats_contain_meanT8_NR10_*.xlsx. We also include results for 50 realisations run on 16 threads in stats_contain_meanT16_NR50

Main GB runs for fixed duration mitigation policies in Imperial College (Report 9 Tables 3 and A1)

• Work in the GB_mitigation folder

• Run batch.ps1 (on Windows) or batch.sh (on Linux) (script will require editing to substitute your favourite batch job scheduler to run on a compute cluster)

• Run summariseMit.r in R (changing cur_path at top of file first)

• Copy first 11 columns of meanT8_NR10\stats_mitigation.csv into the stats sheet of a copy of stats_mitigation_meanT8_NR10.xlsx

• Refresh the pivot tables in other sheets of the spreadsheet

Original results are in stats_mitigation_orig.xlsx. Example results for a range of compilers and platforms are in files stats_mitigation_meanT8_NR10_*.xlsx. We also include results for 50 realisations run on 16 threads in stats_mitigation_meanT16_NR50_msvc.xlsx.

Other notes

• The number of model realisations averaged over can by changed by editing the /NR:10 entry on the command line invocation of the model. Note that results will only match those provided in the xlsx files if the code is run with /NR:10, 8 threads and the current random number seeds.

• The number of threads used by the simulation can be changed by altering the command line option /c:8. A new network file will need to be generated in that case (via the runonce script).

• To output every realisation rather than just the average, change the entry for [Output every realisation] from 0 to 1 in preGB_R0=2.0.txt.

• The last two random number seeds on the command line (which govern the random generated using the epidemic simulation) can be altered without needing to recreate the network file. The first two random number seeds govern generation of the synthetic population and network file. If they are changed, a new network file will need to be generated.