rename tracked property of model.reactions

[sballesteros]

The tracked property of model.reactions seems to create confusion as some users think that the observation model can only observe tracked reactions (whereas it can observe any states including those defined as tracked reactions). Maybe renaming tracked as accumulatedIn will help. @qcaudron feedback ?

[QCaudron]

Agreed. The keyword tracked is misleading. Either accummulatedIn, or perhaps integrated are quite general. For more epi-related context, perhaps just an incidence specifier ?

[sballesteros]

Thanks, trying to stay general and broader than epidemiology so I am afraid that incidence is too epi specific. accumulated seems to be the more general idea. In pomp they use zeronames and "accumulator variables" so I would suggest to go with accumulators (as its a list) or accumulatedIn.

[QCaudron]

Then probably accumulated is good; it's clear, and avoids camelCase. I think zeronames is equally unclear !

[sballesteros]

sorry just edited my comment. accumulators ? Like that it convey that it's a list of states and still avoid the camelCase

[QCaudron]

The property is added to individual reactions. That being the case, adding the property to a reaction one at a time, I'd keep it singular. My vote would be for accumulated, as the reaction is just that - accumulated between timesteps. But it's not hugely important : the major change is to leave tracked behind, I think, to something more intuitively meaningful.

[sballesteros]

See your point. accumulators on the other hand helps to remember that the value has to be a list. Agreed it's not a big deal. @tiffbogich, @JDureau any preferences ? Will change that for the 0.7.x release.

[tiffbogich]

I prefer accumulators to the past tense version. past tense doesn't sounds right to me but i see the point

[JDureau]

I'd suggest sticking to mathematical concepts, hence integrated/tor. I prefer the past form. The -tor sounds cheesy to me.

[sballesteros]

@JDureau 100% agree to stick to the underlying math concepts but here it's more a vanilla sum than an integral no ? Reactions are supposed to be for discrete individuals so we literally cumulate (sum) them. To me integrated involve smtg continuous/infinitesimal (not the case here).

[sballesteros]

That being said, like we will support vanilla ODEs (the model.ode property) in the near future, integrated might be our best bet as it will work everywhere.

[sballesteros]

Although for ODE integrated is weird as every equation is integrated... Maybe accumulators as the ors end clearly indicates that we refer to states variable. I vote for accumulators

[JDureau]

Actually you're opening a nice perspective, allowing integration in the ode or sde objects would lead to differential equations of order > 1. That's really not uncommon. It's far-fetched for the moment, but would be good to keep that door open.

In that case it may be best to avoid potential confusion and step away from integration. What about accumulation? Trying to avoid -or

[sballesteros]

I like the -or as the property describe the creation of new state variables whose job is to accumulate quantities. These new state variables are accumulator variables (that's their mission). Agreed it's a bit terminator kind of crap but I think it's the "right" semantic.

[JDureau]

Ok let's go for it. @tiffbogich and @QCaudron seemed to be fine with that.