Ventana software support forum

The key point for me is that it's easier to build a good model with units than without. Many people regard adding and checking units as an extra step, neglecting the huge savings in effort from avoided errors and improved clarity that it can bring.

It's a somewhat sorry state of affairs. I have to read a lot of unitless papers, because I'm a thread chair. I'd like to summarily reject them in most cases, but that would lead to an unfair asymmetry, because it's easier to judge the quality of a model provided in supplemental material rather than ...

During a game, you can use vensim_get_vecvals to retrieve an array, after first using vensim_get_varoff to retrieve offsets of the variable names. Otherwise, you need to iterate, repeating calls to vensim_get_data for each subscript element for which you want data, i.e. x[id1], x[id2], ... You can u...

Here's one way to do it:

D1 = INTEG( {stuff} , initial D1 )
initial D1 = {a constant}
change in D1 = (D1 - initial D1)

Another option, in DSS, is to define a macro:

:MACRO: INIT(x)
INIT = INITIAL(x)~x~|
:END OF MACRO:

Then you can write

change in D1 = D1 - INIT(D1)

Tom

It's always possible to build a model without units, and adding units won't in itself improve the model. However, in my experience grading hundreds of papers at MIT, it's extremely unlikely that a model will be good the first time. The process of adding units will reveal many conceptual inconsistenc...

I think you need to do something like the following: Treat your drivers as stocks, and your effects as auxiliaries, perhaps with lookup tables to capture any nonlinearity in the effects. Your stocks will have two flows, one associated with an exogenous policy test, and one associated with adjustment...

Actually it's fine to upload screenshots. Like most web fora, you can't copy/paste - you have to save the screenshot locally as a png or jpeg and then use the "upload attachment" tab to include it in your message.

Another option would be to use a tab-delimited file with the GET DIRECT CONSTANTS function (but you'd have to edit the file with something other than Excel. Open Office works, but I'm not sure if it has a column limit).

If your causality runs D1->D2->D3->D1, you have either a feedback loop, or a system of simultaneous equations. If it's a feedback loop, you must think about behavior over time, and there must at least one level (stock) in the loop. If it's a system of simultaneous equations, you can solve it with SI...

Generally, what you're trying to do should be possible. You might check the FIND ZERO example model linked in the help system. I can ponder the details a bit and make further suggestions, I think. Possibly the singular matrix problems arise because the inequality framing admits a multitude of soluti...

You can do basic PDEs by using an array (subscript) as your grid. More sophisticated methods would be tough to implement though - it's not really our focus.

Tom

Hi - The short answer is no. It's possible to have a variable time step, which can sometimes be used for distinct purposes (e.g., to model a dry season and a wet season differently). It's also possible to coordinate two models with different time steps using the .dll, with some kind of coordinating ...

I can't think of a way to do this either. It would be easy working from a programming language, but unfortunately that's not helpful to you!

Check SD conference papers, by googling with the term site:www.systemdynamics.org/conferences/* plus your topic keywords.

Try googling the many SD conference papers on this: site:www.systemdynamics.org/conferences/* GIS i.e. https://www.google.com/search?q=site%3Awww.systemdynamics.org%2Fconferences%2F*+obesity&ie=utf-8&oe=utf-8&aq=t&rls=org.mozilla:en-US:official&client=firefox-a#hl=en&client=f...

A good start is to look at Dennis Meadows' thesis:
http://models.metasd.com/the-dynamics-o ... on-cycles/

Tom

This is possible in DSS/Pro. Are you trying to split a flow into two, or have two sequential valves?

Usually it's cleaner to have separate flows (also from a units and conservation of mass standpoint).

In custom graphs, there's a "dots" option for display. You can also change the regular graph analysis tool to use dots (in Pro or DSS, right-click the tool icon to change settings).

@mike - Right. This is probably the way to go, because Vensim external functions are normally native C, and the DLL is easy to call from C. I haven't used Fortran since the mid-80s, on computers with tape drives, so unfortunately I don't have any further insights on this.

Calibration is just optimization, where the objective is to minimize some measure of the mismatch between model and data. It's always nice when you can hand calibrate the model, but automation is often better and faster. You just need a payoff, specifying which model variables match which data serie...

It's been too long since I worked with user loop functions, but as I recall the main problem occurs when your SubRange has disjoint elements of MainRange; otherwise the call should give you a pointer to the first element of SubRange and you can simply iterate up from there. Is there a reason that yo...

I'm sure you've thought of this, but an extremely simple test is to populate a test argument variable with integers corresponding to the array indices (e.g., x = i), so that your preliminary function can see what it's getting.

Before we get into the details, what is this trying to accomplish it, and why is it impossible in Vensim alone? (There might be an easier way.)

The key difference is in the distribution of arrivals that the two versions imply, so this is really a question to be answered by comparison with data.

You might play with my delay sandbox to get some intuition.
http://models.metasd.com/delay-sandbox/

You can override individual elements - the one that gets overridden is the first active element given the current selections in the Subscript Control.

I don't think there's an easy way to override an arbitrary subset of a variable's elements, without writing equations.