search

tapios / risk-networks

Code for risk networks: a blend of compartmental models, graphs, data assimilation and semi-supervised learning
Other
2 stars 2 forks source link

Do we need node identifiers? #72

Open glwagner opened 4 years ago

glwagner commented 4 years ago

The HealthService may be component even in the absence of node_identifier_helper.py.

@odunbar @lubo93

lubo93 commented 4 years ago

Maybe there is a simpler way, but I used the following code for the Health Service:

health_service = HealthService(patient_capacity = int(0.05 * len(contact_network)), health_worker_population = len(node_identifiers['health_workers']), static_population_network = contact_network)

glwagner commented 4 years ago

Does it work to just use numbers, eg health_worker_population = 20? This is nice, because it shows how to change the number to change the behavior of the program.

lubo93 commented 4 years ago

The number of health workers should be set by "node identifiers". Health workers have degree distribution that is different from other individuals in the network. This is why we need "node identifiers".

odunbar commented 4 years ago

Health workers have degree distribution

I didn't know this... I thought they were like any other node, except they had a certain age group restriction. How is their degree distribution different - surely health workers are people like everyone else? (bearing in mind the HealthService does all of the rewiring online)

glwagner commented 4 years ago

why do health workers have a different degree distribution than other individuals?

glwagner commented 4 years ago

Health workers are recruited here:

https://github.com/dburov190/risk-networks/blob/9b82dabf263c82633c65870b509b76e1e03b4c87/epiforecast/health_service.py#L81

Can we also prescribe the degree distribution of health workers here?

lubo93 commented 4 years ago

Please have a look at the Overleaf document. We use a scale-free degree distribution for the city and an Erdos--Renyi degree distribution for the two remaining groups. It was chosen like that to better match empirical observations.

glwagner commented 4 years ago

How are the health workers connected to the city?

lubo93 commented 4 years ago

How are the health workers connected to the city?

The connection between health workers and city is described by a scale-free degree correction. It's contained in the edge lists that I generated.