Handle missing values in block decomposed version of Kalman filter

Mostly implemented as of 16e9c36e.

Still needs to be documented and polished.

Also need to implement the filter with missing values (for GPM)

assigned to @sebastien

changed milestone to %5.x

added documentation label

@JohannesPfeifer I actually have no plan to implement this. Also, why did you assign it to 4.7?

unassigned @sebastien

My understanding was that only the documentation remains to be done, which should be feasible for 4.7.

I see, that wasn’t clear from the ticket history. But what exactly is there to document then? Just to say that it works? AFAIR, there is no dedicated option.

I don't know. You wrote above "needs to be documented and polished". !1911 (merged) was the relevant bugfix. If you think there is nothing to be done, we can close the ticket.

My guess is that this ticket (imported from our old Trac instance) was initially about the block decomposed Kalman filter in general, then later got renamed to “Handling missing values” when the main filter was merged. But that renaming information (between my first and second comments) was lost in the import process (to Github). Hence the false impression that missing values just have to be documented.

As you can see from line 701 of matlab/dsge_likelihood.m, the block Kalman filter is never called when there are missing data.

So I think this was never implemented.

removed milestone %5.x

removed documentation label