🎼 Master B major to control F major: simulation of, by, for Box-Flow

[hyunjimoon]

Are we confusing MICRO and MACRO world?

I think we are applying probability sum law (mutual exclusive) and multiplication law (independent) in the same place but shouldn't..

"AR = Adoption from word of mouth + Adoption from advertisement"

Hazhir explained in infinitesimal time, one person cannot adopt both from advertisement and mouth. However, why does 'one person' view appears all of a sudden?

And might this be relevant to problem "p4 which is a mixture is not the same as p3" Holes in Bayes related to quantum physics described below?

box-flow-box VS flow-box-flow?

Flow_major: faster = open system = flow-box-flow = information, soul (move) = well-mixed = micro (= better with compartment model?)
Box_major: slower = closed system = box-flow-box = material, body (regulate) = less-mixed = macro (= better with agent model?)

"Jules Verne: The body regulates the soul, and, like the balance-wheel, it is submitted to regular oscillations."

In the following model Repair_Inven,

Repair is a Box_major and markov chain while Inven is Flow_major. When the two are put in one model, Repair seems to be ever-lasting as it is compared against Inven which has shorter time-horizon (i.e. or may I say less sustainable? not that equilibrium is impossible in Inven system, but compared to Repair system it is. Hazhir mentioned finding equilibrium in Project model is useless because that would mean the project has ended and I think it might be relevant..)

However, I am no sure yet how to understand the two different formulation of SIR model (above Tom's, below: Hazhir's)

More explanation on box-flow-box vs flow-box-flow and openness of system in https://github.com/Data4DM/BayesSD/blob/master/DailyDigest/4%20System%20Structures.md#3-openness-informationenergymaterial

[hyunjimoon]

+ is a symbol

In a higher probability space "+" is not the same any more. Fuzzy min and max.

• Transition between Discrete and Continuous time Markov chain is not straightforward.
  Compartment models is mixing MICRO_WORLD (rate addition (based on independence) with MACRO_WORLD (integration, ratio

• Deterministically non-existing but stochastically existing is my conclusion.. Much information is lost upon wlog z1 > z2 which can be overcome with distributional equivalence in stochastic design

Discrete and Continuous

discrete and continuous [code](https://github.com/hyunjimoon/defense-reliability/blob/17ad362d92e91f508ada9062b9f2a4ae5c2cd01e/twin/RepairInven.ipynb)

[tomfid]

Not sure what the minimax thing is about, but the CES production function is a soft-min that approaches the min function in the limit. https://math.stackexchange.com/questions/604216/how-ces-function-becomes-minimum-function

[hyunjimoon]

Functional formulation, coflow (how to overcome the independent assumption or full mixing), allocation may be relevant of rainbow between C and A (will call California rainbow :))
Hazhir has a paper

System Dynamics Review - 2010 - Rahmandad - Modeling the rework cycle capturing multiple defects per task.pdf

[tomfid]

Lots of questions here.

I think the "one person" perspective arises because a person is the natural atomic agent in this system. The additivity is a consequence of the standard DES/queuing result that any sufficiently short interval has vanishing probability of more than one Poisson event.

However, that's contingent on the simple model of causality depicted in the Bass model: a contact with a user or ad instantly induces conversions (with some probability). If you built an agent version of the model with individuals motivated in that way, and well-mixed contacts, you should get exactly the same result.

I'm sure it's possible to construct a different model of the individual, where for example P(conversion) increases in a nonlinear way with accumulated exposure. That would potentially be much harder to aggregate. OTOH in a lot of models, the difference might not be important.

[hyunjimoon]

@tomfid from Nathaniel's "characterizing heterogeneity in agent based modeling" , the following slides caught my eyes

[hyunjimoon]

[hyunjimoon]

Sustainability targets F major, but B-major should be mastered in advance:
i) capability
ii) perceived drum for sync signal of system components

[hyunjimoon]

From 879 seminar models here I sorted by the number of box + flow.

model no.		sum	box	flow
1	SEIRb	11	6 suceptible, exposed, infected, recovered, dead, death_perceived	5 exposing, onsetting, recovering, dying, death_perceiving
2	project	12	6 project_resource, task_to_be(=backlog)-{done, reworked, quality_assuranced} , task_released, project_lasted	6 resource_adjusting, completing, rework-{discovering,doing}, work_approving, resource_adjusting, dealine_delaying
3	market growth	12	6 sales_force, recent_revenue, backlog, delivery_delay_market_preceived, delivery_delay_company_preceived, capacity	6 salesforce_net_hiring, revenue_net_changing, backlog-{ordering,shiping}, capacity_net_acquisiting, delivery_delay_perceiving
4	long wave	12	5 order_expected, capital-{supply_line, stock}, goods_supply_line, backlog	7 capital-{ordering,acquisiting,discarding}, capital_backlog-{ordering,producing}, goods_capital-{ordering,acquisiting}
5	mat_info_price	17	8 production_cost_expected, industry_demanded, {long, short}-run_price_expected, variable_cost_expected, capacity_utilization, wip, inv	9 product-{starting, ing, shiping}, charm-{{capacity_util, industry_demand}-nonprice, {short, long}-price}-adjusting, cost-{expected, variable}-adjusting
6	service quality	27	13 labor_desired, labor_productivity_perceived, vacancies, rookies, personnel_experienced, backlog_service, desired_to, fatigue-{A, E}, cust_service_expected, employee_quality_expected, mgmt_quality_perceived, cust_quality_perceived	7 + 7