Jim hynes' three wishes 1. Better tools for model analysis, meaning making it easier to trace a particular aspect of behavior to structure and understanding how the structure generates that aspect of behavior. 2. Automatically increase robustness of a model -- i.e. make the modeled hypothesis robust to parameter changes and simple structural change (e.g. putting in a perception delay). 3. Help creating policy (decision rule) changes. Perhaps changing a single answer (e.g. introduce a product at this point in time) to a policy answer. | python function | purpose | in: xarray `data_vars` | out: plot | required work (priority) | | ------------------ | ------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------- | ----------------------------------------- | ---------------------------------------------------------------------- | | `tf11234_vgcsv2xr` | run vengine and store csv in `xarray` | | | refactor (1), run vengine for 3x23 csv files (2) | | `tf5_x_ctnt` | are real and simulated cost component time vector of three main states similar? | cost and vaccine taking population, aggregated over space `x_ct`, `x_nt` | ![[Pasted image 20230709113704.png\|300]] | automate vengine-csv (real,sim of gdp,death,hosp.of three states) (5) | | `tf6_f1x_cs` | how does each state's cost component (gdp, death, hosp.) differ? | last day's cumulative mvv calculation for baseline `f1x_pcst[{'param':'baseline', 'time':time[-1]}]` | ![[marg_vv_by_state2.png\|300]] | refactor (4) | | `tf7_f2x_p` | which parameter affects cumulative mvv the most? | sensitivity of last day's cumulative mvv w.r.t. 22 parameters`f2x_p` | ![[elasticity_marg.png\|300]] | refactor (4) | | `tf8_x_ch2` | how would vaccine timeshift affect cumulative mvv and its each component? | `x_ch2` | ![[vv_by_vac_start_offset.png\|300]] | refactor (3) | | | | | | | expected time: 1,3,4: 3hr + 2: 3hr (+alpha)