Matthieu Gomez

I am not sure I understand everything, so I am sorry if I misunderstand you, but my point is that the HJB is not a linear operator in my example,...

Yes, your description of the fully implicit (EconPDE) method vs semi implicit (HACT) is exactly right. I fully agree with your tradeoff convergence vs speed. The fully implicit method gives...

Thanks @ChrisRackauckas. That is very interesting. Thanks for the links, I will look more into them. The implicit method works well in the economics problems I have worked because if...

Not related to this particular user case, but I'd love it to see `missing` values encoded by a reference of zero. It makes it easier to obtain efficient code since...

Any thought?

I’m not sure it would be that ideal to have a lazy implementation of this lag function — no one wants a vector that is as slow as a dictionary....

How would that work with panel data though? ie using lag on a grouped DataFrame? On Mon, May 4, 2020 at 3:05 PM Milan Bouchet-Valat wrote: > To me it...

I don’t think that would work: time stamps of a TimeArray need to be unique. On Mon, May 4, 2020 at 3:31 PM Milan Bouchet-Valat wrote: > Well GroupedDataFrame gives...

Yes, exactly.

I don’t think it would work since you may want to lag by more than one period. Actually, it looks like TimeArray actually has a lag function, which does not...