Select variables for variance-based sensitivity analysis

bwbush commented 4 years ago

Here are the candidates from the elementary-effects analysis: I've checked the variables whose median rank for influence is in the top 30 for mu* or sigma. Please check or uncheck variables to alter the selection for the variance-based study. See also the diagrams below, which show that some input variables are influential for just a couple of output variables, but others are broadly influential.

[x] random stream
[x] target pilot hours
[x] base external investor ask rate
[x] commercial plant capacity input
[x] pilot capacity
[x] commercial capital cost input
[x] piloting acceptable rate
[x] management response time
[x] commercial variable operating cost input
[x] payback period multiplier
[x] minimum runway
[x] pilot and demo response time
[x] startup piloting rate
[x] commercial process yield input
[x] piloting failure default recovery time
[x] bioproduct long term price
[x] researching impact on piloting
[x] government capital cost share
[x] strategic value to external investors
[x] piloting failure distribution max
[x] startup piloting period
[x] commercial plant capacity
[x] expected continuity of government policy
[x] government production incentive
[x] aversion to NPV deviation
[x] required return multiplier
[ ] commercial fixed operating cost input
[x] bioproduct performance advantage
[x] target demo hours
[ ] business as usual runway
[ ] piloting management overhead
[ ] demo capacity
[x] initial market size
[ ] bioproduct price reversion time
[ ] regulatory process starting point
[ ] pathway approval required switch
[ ] researching impact on demoing
[ ] investor optimism
[ ] startup demoing rate
[ ] demoing failure default recovery time
[ ] bioproduct price fluctuations
[ ] duration of intellectual property
[ ] investor history
[ ] product approval maximum time
[ ] product approval required switch
[ ] demoing acceptable rate
[ ] demoing failure distribution max
[ ] incumbent maximum fluctuation
[ ] startup demoing period
[ ] expected green premium
[ ] elasticity of demand
[ ] feedstock approval required switch
[ ] incumbent reversion time
[ ] long term market size
[ ] max management runway response
[ ] bioproduct offtake agreement
[ ] government research cost share
[ ] incumbent long term price trend
[ ] advertising start time
[ ] incumbent market share target economic
[ ] incumbent price response magnitude
[ ] incumbent starting price
[ ] pathway approval maximum time
[ ] demoing management overhead
[ ] feedstock approval maximum time
[x] market growth rate
[ ] advertising budget
[ ] effectiveness of intellectual property
[ ] retrofit delay
[ ] government operating cost share
[ ] elasticity of supply
[ ] commercial plant startup period
[ ] custom feedstock long term price change
[ ] custom feedstock maximum fluctuation magnitude
[ ] custom feedstock periodic fluctuation magnitude
[ ] custom feedstock reversion time
[ ] custom feedstock starting price
[ ] feedstock approval maximum cost
[ ] government operating grant period
[x] number of missed stagegates allowed
[ ] pathway approval maximum cost
[ ] product approval maximum cost
[x] required internal return
[ ] stagegate length

Heat map of mu* ranks for input variables.

Box plots of mu* ranks for input variables.

Heat map of sigma ranks for input variables.

Box plots of sigma ranks for input variables.

bwbush commented 4 years ago

Note that the elementary effects started with the generic template.cin, so some variables might not be active in this base case even though they'd be active in other base cases. An example is required internal rate of return. In principle, we should perform an elementary-effects study for each base case.

Lauren-Sittler commented 4 years ago

Based on the heat maps, and which variables I thought would be interesting, I selected: target demo hours, bioproduct performance advantage, initial market size, and market growth rate. I also selected required internal return, and number of missed stagegates allowed. I am concerned that those two variables are showing as non-influential because the internal investment switch is off. The base external ask rate is very influential, so I'm wondering what would happen if that switch were changed to an internally funded project.

NREL / BioproductTransitionDynamics