Epidemiological Parameters

See also

Default Model Example

epi_params.py

Calculate delay distributions and generate delay parameter dictionaries for model building.

class epimodel.pymc3_models.epi_params.EpidemiologicalParameters(seed=0, generation_interval=None, incubation_period=None, infection_to_fatality_delay=None, infection_to_reporting_delay=None)

Bases: object

Epidemiological Parameters Class

Wrapper Class, contains information about the epidemiological parameters used in this project.

discretise_samples(samples, max_int)

Discretise a set of samples to form a pmf, truncating to max.

Parameters

  • samples – Samples to discretize.

  • max – Truncation.

Returns

pmf - discretised distribution.

generate_dist_samples(dist, nRVs, with_noise)

Generate samples from given distribution.

Parameters

  • dist – Distribution dictionary to use.

  • nRVs – number of random variables to sample

  • with_noise – if true, add noise to distributions, else do not.

Returns

samples

generate_gi(with_noise=True)

Generate generation interval parameters.

Note: this uses the random seed associated with the EpidemiologicalParameters() object, and will be consistent.

Param

with_noise: boolean - if True, add noise to estimates

Returns

Mean, sd of generation interval.

generate_pmf_statistics_str(delay_prob)

Make mean and variance of delay string.

Parameters

delay_prob – delay to compute statistics of.

Returns

Information string.

generate_reporting_and_fatality_delays(nRVs=10000000, with_noise=True, max_reporting=32, max_fatality=48)

Generate reporting and fatality discretised delays using Monte Carlo integration.

Note: this uses the random seed associated with the EpidemiologicalParameters() object, and will be consistent.

Parameters

  • nRVs – int - number of random variables used for integration

  • max_reporting – int - reporting delay truncation

  • with_noise – boolean. If true, use noisy values for the incubation period etc, otherwise use the means.

  • max_fatality – int - death delay trunction

Returns

reporting_delay, fatality_delay tuple

get_model_build_dict()

Grab parameters which can be conveniently passed to our model files.

Returns

param dict

summarise_parameters()

Print summary of parameters.

epimodel.pymc3_models.epi_params.bootstrapped_negbinom_values(delays, n_bootstrap=250, n_rvs=10000000, truncation=64, filter_disp_outliers=True)

Fit negative binomial to n_bootstrapped sets of n_rv samples, each set of samples drawn randomly from the priors placed on the distributions in the delay array. e.g., this function is used to fit a single negative binomial distribution (with uncertainty) to the sum of the incubation period and onset to death delay.

Parameters

  • delays – list of distributions (with uncertainty), used to produce estimates.

  • n_bootstrap – number of bootstrapped to perform

  • n_rvs – number of samples to draw from each draw from prior

  • truncation – maximum value to truncate to.

Returns

dictionary with uncertain NB values

epimodel.pymc3_models.epi_params.ci_to_mean_sd(mean, ci, percent=0.95)