SIR results don't make a ton of sense

[cliffckerr]

Modifying the parameters of the SIR model, it's not behaving the way I'd expect. Some preliminary findings:

• [ ] Initial prevalence probably shouldn't be counted as a huge spike of infections at t=0
• [ ] Hard to interpret death rates. p_death=0.5 results in a handful of deaths out of 10k people, but p_death>1 leads to an exception
• [ ] prevalence isn't being calculated incorrectly -- like people aren't being removed from n_infected when they recover?
• [ ] Beta seems non-intuitive to set, hard to get a typical SIR-shaped curve

[cliffckerr]

Closed by #330

[kaa2102]

Update: This specifically references the simpleabm code via the agent_model.py file:

I think the issue in agent_model.py is that the Susceptible population should be declining at a faster rate per standard SIR models.

`# Run the simulation for t in x[:-1]:

pop.check_infections() # Check which infectious occur
pop.check_recoveries() # Check which recoveries occur

#S[t+1] = pop.count_S() # Count the current number of susceptible people
I[t+1] = pop.count_I()
R[t+1] = pop.count_R()

# --> Update Susceptible Math <--
S[t+1] = N - I[t+1] - R[t+1] # Population - sum(Recovered) - sum(Infected*)
#*If Infected confers immunity or exclusion from the Susceptible population.

`

[kaa2102]

The tests/test_simple.py has a test_sir_epi() method method that generates objects: s0, s1 ->s3a, s3b (output in main)

Potential offending method: cum_infections() where s3a Infected array has all zero '0' values and s3b has incorrectly calculated Cumulative Infections. https://github.com/amath-idm/starsim/blob/a88795b0b39693b3e25b0f2d082d7d258f3d1b5d/tests/test_simple.py#L75C1-L86C52

Code in disease.py

    def update_results(self, sim):
        super().update_results(sim)
        res = self.results
        res['prevalence'][sim.ti] = res.n_infected[sim.ti] / np.count_nonzero(sim.people.alive)
 >>>res['new_infections'][sim.ti] = np.count_nonzero(self.ti_infected == sim.ti) 
        res['cum_infections'][sim.ti] = np.sum(res['new_infections'][:sim.ti])

np.count_nonzero() will produce an integer value for the count of nonzero array elements given an array input of numbers (integer and/or float). However, self.ti_infected == sim.ti this looks to evaluate as a boolean logical expression with output as an array of boolean values of True, False = 1, 0. I am not 100% if self.ti_infected and sim.ti are both arrays. This *may indicate why s3a generates Infected array values all equal to zero where all expressions evaluate as False.