Upgrading ST-Sim from Version 2 to Version 3 – Important Changes

[leonardo-frid]

I recently had an e-mail from a model user who found they had very different results in non-spatial simulations between ST-Sim versions 2.4 and 3.0. This change is related to the way transitions are simulated non-spatially. In this post I explain the reasons behind this change and provide instructions on how to configure non-spatial models in version 3.0 and later so that results match those of version 2.4.

In ST-Sim version 2.4 and earlier there was a major discrepancy between non-spatial and spatially explicit simulations because the former only allowed a single transition per cell per timestep whereas the latter allowed multiple transitions (one per primary transition group) per cell for each timestep. In order to remove this discrepancy we now allow multiple transitions in non-spatial model runs (one per primary transition group). Prior to this change the concept of primary transition groups was unnecessary in non-spatial model runs. However, now if a transition is assigned to more than one group, users must specify which of these groups is the primary one which will be used at runtime. If users do not specify, the model will assume all groups as primary and simulation cells may experience the same type of transition multiple times (once for each group that it belongs to). If you have transition types belonging to multiple groups, the easiest way to ensure that a non-spatial model run will behave exactly the same in version 3 as in version 2 is as follows:

1. Under Project definitions | Transitions | Transition Group add a single transition group called All Transitions
2. Under Project definitions | Transitions | Transition Types by Group right click on the table to show the Is Primary field.
3. Still in Project definitions | Transitions | Transition Types by Group set the Is Primary field to No for all existing records
4. Still in Project definitions | Transitions | Transition Types by Group add a record for each transition type assigning it to the All Transitions group and setting the Is Primary field to Yes for these records only.

Your model run should now behave the same as in version 2.0 with only one transition possible per cell.

[leonardo-frid]

A followup e-mail I received on this topic:

One clarifying question I have now is: given that only one transition can happen per cell per time step, what determines the order of operations? Which of the transitions has a chance to happen first, second…. and possibly exclude other transitions from taking place?

Within each primary transition group the non-spatial model iterates over all cells and and for each of these considers simultaneously which, if any, of the possible transitions for that primary group will occur. The algorithm is such that there is no bias towards one transition group excluding another. In the most recent version of ST-Sim, users have the ability to specify the order in which they would like the model to iterate across primary transition groups. The order can be specified as iteration and time varying. If the order is unspecified, the model randomizes the order each timestep.

[Author]

Posts