Testing new stanify - replicated? datastructure?

[hyunjimoon]

@Dashadower compare_vensim_stan.mdl.zip below is a vensim file with the datastructure you've requested -- Light blue variable exog demand is datastructure and dark blue table for order fulfillment ratio is the table function. Could you please check the below?

1. setting value for exog demand makes this variable no longer datastructure. If my memory is correct, you were trying to receive data only from vensim. However, I don't know how this can be implemented other than the original method (supplying value in python after declaring datastructure in vensim). Happy to be corrected!

2. time series patterns are replicated in the result from stan's draws2data. For easy comparison, I've set exog demand is 2 * Time + 100 (this requires changing exog demand type (so it's not datastructure anymore ) as shown below.

With the value above, please make sure the overshoot of adj ss and smooth adjustment of adj s are visible in the stan-generated time series.

compare_vensim_stan.mdl.zip

[Dashadower]

In the Vensim model you've provided, I couldn't find any instances of exog demand being used with a supplemented argument. Is exog demand supposed to be a single scalar value?

[Dashadower]

In addition, I request that you provide the Vensim file with exog demand changed to an expression. It's hard to modify Vensim files without an actual Vensim client.

[hyunjimoon]

I couldn't find any instances of exog demand being used

I don't follow; don't you have the following in the zipped mdl file?

EOR= INTEG (
	adjusting eor,
		exog demand)
	~	widget/Week
	~		|

B in=
	exog demand
	~	widget/Week
	~		|

[Dashadower]

Nah, it's there. It's just that exog demand is being used as a static scalar value; I'm not sure when you would generate a function for it like lookup.

[hyunjimoon]

From below, exog demand's type is datatype and time series vector is supplied (perhaps through python @ in stanify). in vensim, I recall we implemented as interpolate/lookup structure. Does thie make sense to you? Is everyting good?

exog demand:INTERPOLATE:
~ widget/Week
~ |

[Dashadower]

Then is the default assumption for Vensim data variables is they exist as a time-vector?

[hyunjimoon]

yes