sis_bpf: Bootstrap Particle Filter for SIS model with Population...

In nianqiaoju/agents: Toolkit for inference in agent-based modelss

Description

approximates marginal likelihood of the SIS model using bootstrap particle filter NOTE: this cannot do predictions yet.

Usage

sis_bpf(y, model_config, particle_config)

Arguments

y	a vector of length (T+1)
model_config	a list containing: 'N': size of the population 'alpha0' : initial infection probability 'lambda': infection rate 'gamma': recovery rate 'adjacency_matrix_b': network structure 'rho': reporting rate
particle_config	a list containing: 'num_particles' : number of particles for the bootstrap particle filter 'ess_threshold' : if effective sample size drops below the threshold, then perform a resample step. ess_threshold = 1 means resampling at every step. 'save_particles': binary 'clock' : binary, default to FALSE. If clock = TRUE, then we will use a stopwatch to document its Sys.time() 'save_genealogy': binary 'verbose': if TRUE print messages for debug

Value

A list containing

• 'log_incremental_likelihood' :

• 'log_marginal_likelihood' :

• 'log_final_likelihood' :

• 'particles': [N, (T+1), num_particles] array storing the particles:

• 'eves' : the unique ancesters of each particle in the genealogy structure, if save_genealogy = TRUE;

• 'ess' : effective sample size at each step

• 'runtime' : elapse time at each step, if clock = TRUE;

• 'totaltime' : elapse time of running the whole particle filter, if clock = TRUE;

nianqiaoju/agents documentation built on Feb. 19, 2021, 12:18 a.m.