Change in general IPM vignette

[Aariq]

I noticed you recently fixed a mistake in the General IPM vignette (b6af232) and I changed the General IPM I'm working on accordingly. To my surprise, the output of that IPM changed. I'm just curious what it was doing before when it was trying to integrate a discrete class over a state variable it didn't have. I wasn't expecting that fix to have any effect. It might be a good idea to add a check if a kernel is discrete (family = "D_") and throw an error if it has a d_<state var> in the formula.

btw, that vignette is soooooo helpful 🙏🏼.

[levisc8]

There should be a change, as the integration with midpoint rule just multiplies the expression by the value of the bin width (i.e. d_z). In this case, I think it's a pretty small number.

As for the broader checking, what I really need to do is just remove the need for users to supply d_zs to begin with. Early on, I didn't fully understand how the integration was functioning in the model (i.e. always integrating w/r/t to initial states). In short, if you're starting on "D", then no integration, and if you're starting on "C", then it should be d_<state_start>. It's a fairly straightforward thing to implement, I just haven't gotten around to doing it yet.

I'd probably keep a check for d_z for backward compatibility - it'd suck if this update for convenience broke a model you'd already made! Hopefully will have time for this in the not-so-distant future!

PS Sorry that the discussion tab hasn't taken off!

[Aariq]

Thanks for the quick reply!