integration rules

[levisc8]

New options should be "cdf" and "b2b".

It may make sense to re-examine how expressions are evaluated internally now. I think pretty much all vital rates will now have to become call2(<int_rule_fun>, !!! vr_expr) where <int_rule_fun> sets up the meshes and modifies vr_expr appropriately.

e.g. "cdf" needs to change Norm(size_2, mu_g, sd_g) to pnorm(size_2_u, mu_g, sd_g) - pnorm(size_2_l, mu_g, sd_g)

[wpetry]

I can help with this. Can you point me to where this is?

[levisc8]

Hi Will, sorry for the slow response. Unfortunately, I think this points to a larger design flaw in the package itself - midpoint integration rule isn't implemented anywhere so much as it's baked into the syntax.

Implementing the midpoint rule should be straightforward enough - i think you'd just write a wrapper called midpoint() that evaluates the expressions just as they currently are evaluated.

The cdf approach would require an additional translation layer built into it (maybe see here for an example of what that might look like). However, I'm not sure what step in the make_ipm process is best to implement that translation.

I suspect the b2b approach would require this translation as well, though I have to confess that I've forgotten what that approach actually does (haven't thought about IPMs for a little while now!)