Individual Sensitivity Experiments¶

Agg Holdout¶

Holdout the final 20 days of data.

usage: python agg_holdout.py [-h] [--model_type MODEL_TYPE] [--exp_tag EXP_TAG] [--n_chains N_CHAINS]
                             [--n_samples N_SAMPLES]

Named Arguments¶

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

Alternative Priors¶

Alternative priors for basic reproductive number R0 and NPI effectiveness.

usage: python scripts/alternative_build_param.py [-h] [--R_prior_mean R_PRIOR]
                                                 [--NPI_prior NPI_PRIOR NPI_PRIOR]
                                                 [--growth_noise GROWTH_NOISE] [--model_type MODEL_TYPE]
                                                 [--exp_tag EXP_TAG] [--n_chains N_CHAINS]
                                                 [--n_samples N_SAMPLES]

Named Arguments¶

--R_prior_mean

Prior mean basic reproductive number R0

--NPI_prior

Prior for NPI effectiveness

--growth_noise

Growth noise scale parameter

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

Any NPI Active¶

Add an additional dummy NPI representing whether any major NPI is active

usage: python scripts/any_npi_active.py [-h] [--model_type MODEL_TYPE] [--exp_tag EXP_TAG]
                                        [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

Delay Schools¶

Add artificial 5 day delay to school closure NPI.

usage: python scripts/delay_schools.py [-h] [--model_type MODEL_TYPE] [--exp_tag EXP_TAG]
                                       [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

NPI Leaveout¶

Remove NPIs from the dataset, fitting data to the remaining NPIs.

usage: python scripts/npi_leaveout.py [-h] [--npis NPIS [NPIS ...]] [--model_type MODEL_TYPE]
                                      [--exp_tag EXP_TAG] [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--npis

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

NPI Timing¶

Replace NPIs with indices 0-8 representing 0,…,8 active NPIs, ignoring NPI type

usage: python scripts/npi_timing.py [-h] [--model_type MODEL_TYPE] [--exp_tag EXP_TAG]
                                    [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

OxCGRT Leavin¶

Include additional NPIs from OxCGRT

usage: python scripts/oxcgrt_leavein.py [-h] [--npis NPIS [NPIS ...]] [--model_type MODEL_TYPE]
                                        [--exp_tag EXP_TAG] [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--npis

OxCGRT NPIs to include. One or more of:: Travel Screen/Quarantine

Travel Bans

Public Transport Limited

Internal Movement Limited

Public Information Campaigns

Symptomatic Testing

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

Preprocessing Tests¶

Modify the smoothing window, threshold number of cases or threshold number of deaths below which data is masked.

usage: python scripts/preprocessing_tests.py [-h] [--smoothing SMOOTHING]
                                             [--cases_threshold CASES_THRESHOLD]
                                             [--deaths_threshold DEATHS_THRESHOLD]
                                             [--model_type MODEL_TYPE] [--exp_tag EXP_TAG]
                                             [--n_chains N_CHAINS] [--n_samples N_SAMPLES]

Named Arguments¶

--smoothing

Number of days over which to smooth. This should be an odd number. If 1, no smoothing occurs.

--cases_threshold

Deaths threshold, below which new daily deaths are ignored.

--deaths_threshold

Confirmed cases threshold, below which new daily cases are ignored.

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

Structural¶

Alternative model structures.

usage: python scripts/structural.py [-h] [--model_type MODEL_TYPE] [--exp_tag EXP_TAG]
                                    [--n_chains N_CHAINS] [--n_samples N_SAMPLES]
                                    [--model_structure MODEL_STRUCTURE]

Named Arguments¶

--model_type

model structure choice:: - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i)

- discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval

- discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals

- noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion)

- different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c}

- cases_only: the number of infections is estimated from case data only

- deaths_only: the number of infections is estimated from death data only

--exp_tag

experiment identification tag

--n_chains

the number of chains to run in parallel

--n_samples

the number of samples to draw

--model_structure

model structure choice: | - additive: the reproduction rate is given by R_t=R0*(sum_i phi_{i,t} beta_i) | - discrete_renewal_fixed_gi: uses discrete renewal model to convert reproduction rate R into growth rate g with fixed generation interval | - discrete_renewal: uses discrete renewal model to convert reproduction rate R into growth rate g with prior over generation intervals | - noisy_r: noise is added to R_t before conversion to growth rate g_t (default model adds noise to g_t after conversion) | - different_effects: each region c has a unique NPI reduction coefficient alpha_{i,c} | - cases_only: the number of infections is estimated from case data only | - deaths_only: the number of infections is estimated from death data only