Notations on transitions

[choisy]

I'm confused again about the first model in the README. In particular, if I understand well, when you use mathexpr(), the distribution that is implicitly assumed is an exponential one right? In that case, that would be much nicer if we could write

"V -> IV" = exponential(beta * 0.3 * V * (I + IV) / N)

instead of

"V -> IV" = mathexpr(beta * 0.3 * V * (I + IV) / N)

And, again, I don't think we need mathexpr() and constant() in this package.

[thinhong]

The reason I made the mathexpr in C++ was because I needed something to calculate with the number of people in >1 compartments (in this example we have 3 compartments V, I, IV), and if we want something more than just addition, substraction, multiplication... You're right in R we don't need a function for it, I can check if the object is a string then make it a mathexpr in C++.

What I understand from the exponential distribution is when we have a constant rate, i.e when the number going out of V is only beta * V then it is exponential, could you confirm? Anyway, in C++ it has to be handled by a mathexpr because there are >1 compartment and not only the left compartment involved lol.

Actually, I'm thinking about dropping the exponential() because when I set the probability() it will be the same as the exponential() as this is discrete time?

[thinhong]

I meant because the I and IV keep changing through time so beta * (I + IV) is not constant through time, and the length of stay in V is not exponential?

[choisy]

I meant because the I and IV keep changing through time so beta * (I + IV) is not constant through time, and the length of stay in V is not exponential?

Yes, good point, you're right on this one. But please get rid off mathexp(), it's ugly. It's not because you need it in the C++ code that it should appear in the R interface. The whole point of the R interface is to hide as much as possible the technicalities of the C++ code. Thus, we want this user interface language as slim as possible.

[choisy]

Not sure what you mean about probability() here. Are you talking about transprob() by any chance? If yes, I think that's another good point on which you're right again, in which case the diagramme would now look like this right?