Question from meeting with @tamuri, @giordano, @matt-graham and @willGraham01

• Do we want to do this on a single (larger) machine along with mode 1 runs or a separate machine?

Seems like we're leaning towards former.

@marghe-molaro / @tbhallett - we're working on adding mode_appt_constraints=2 to the nightly long runs. Couple of questions:

1. If we're using the long_run_all_diseases.py scenario file to run the mode=2 runs, is the only necessary change overriding the mode_appt_constraints parameter? This parameter is set to 1 by get_parameters_for_status_quo() but we now have a way to change it to 2 on-demand.
2. Will the post-processing scripts (that create the plots) require any modification?

Hi @tamuri,

1. I think the "dream" nightly long runs scenarios for mode 2 would be as follows:

                "HealthSystem": {
                    "mode_appt_constraints": 1,                            # <-- Start in mode 1
                    "mode_appt_constraints_postSwitch": 2,      # <-- Include a transition to mode 2, to pick up any issues with this
                    "year_mode_switch": 2012,                              # <-- Could make this quite soon, but I'd say >1 year
                    "tclose_overwrite": 1,                                        # <-- In most of our runs in mode 2, we chose to overwrite tclose
                    "tclose_days_offset_overwrite": 7,                  # <-- and usually set it to 7.
                }

Especially the latter two, however, are not crucial. If it were much simpler to run in a simple mode_appt_constraints =2 I think it would be fine.

2. I think maybe a time evolution of HSIs delivered, either as total number or fraction of requested that are delivered, and a time evolution of DALYs. Both could be done for the case of HIV/AIDS only, which are the main drivers of HSIs and DALYs initially.

@tbhallett, what do you think?

1. that seems fine to me. (If that's the target for most of the runs we use, which I think it is)

2. I think the original processing scripts can be used but it sounds like - in addition - Margherita would like to see the time trends in certain things which the original scripts don't look at.

I think the run is only from 2010-2019, and the target period for calibration is 2010-2019.

Perhaps, as the main purpose of this is to flush out bugs when running in this mode, we can keep the original scripts for now, and we can later raise a PR that brings in additional scripts to look at time trends.

Is that ok, Margherita -- or do you think longer runs now and new scripts for analysis would be necessary for the first step?

Is that ok, Margherita -- or do you think longer runs now and new scripts for analysis would be necessary for the first step?

Yes I agree!

To @tamuri's point above, we could make a function get_parameters_for_mode2_run or something like that, and put that specification in there. That way we can edit it easily down the line....?

Ah sorry I missed that, yes I can issue a PR for a get_parameters_for_mode2_run function (once I've submitted the module order tests)

Update: this now included in PR #1398