Allow the user to prescribe mitigation date

[PhilMiller]

This is on top of various cleanup and refactoring in #415, so probably review that first.

There's a potential edge case if distancing is strong enough to cause a dip in the census curve, that prescribed-doubling fitting of first-hospitalization might align to a point on either side of the dip.

I've tested this manually and it seems to behave as expected, but I don't have a great handle on how to really validate it.

Fixes #407

[PhilMiller]

Documentation and input description wording probably need work. As implemented, without #340 taken into account, the date given needs to be when effects of distancing or other mitigation were/will be seen, not the date a policy was enacted.

[PhilMiller]

Playing with things more, I realize that the fit could potentially have three points where the given census crosses the curve in the allowed time before the forecast maximum

Again, I believe this is only relevant if the user sets the doubling-time rather than the date of first hospitalization.

[PhilMiller]

Heh, I was actually thinking about how we ought to make n_days go away as well. I agree on the passing a list of pairs, and of not making that class state. I was trying to get the feature built with minimal other disruption, but if you want the further refactoring, I'm onboard to do it.

…

On Wed, Apr 1, 2020, 22:22 Jason ***@***.***> wrote: ***@***.**** commented on this pull request. ------------------------------ In src/penn_chime/models.py <#418 (comment)>: > @@ -147,17 +147,29 @@ def __init__(self, p: Parameters): self.daily_growth_rate_t = get_growth_rate(self.doubling_time_t) def run_projection(self, p): I might pass the (beta, n_day) pairs into run_projection(p, *args), instead of using state. The call site knows better if it is pre or post fit, and how many of these pairs are needed. If I was going to use state, then I would store the pairs as in a list: self.policies = [beta1, n_days1, beta2, n_days2, beta3, n_days3, beta4, n_days4...] ... self.raw_df = sim_sir_df( self.susceptible, self.infected, p.recovered, self.gamma, -self.i_day, *self.policies) ... to avoid having to rewire run_projection when extending the ui to allow multiple (social_distancing, date) pairs. Ideally, the n_days input from the UI input would be merely a UI setting for the graphs, and gen_sir would simply run until i < 0.5. This would avoid complicating the model total days computations — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#418 (review)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAA64ABEP5E634KY7SAKNSTRKPZHHANCNFSM4LZQ3SNA> .

[jlubken]

Thanks @PhilMiller !

[misken]

Is the new mitigation-date parameter going to make it into the CLI?