Markov solver in DCL-Dev-1 about hark HOT 5 CLOSED

Did I have cubic interpolation in my version? I know there was a bit of new
math required, and I vaguely remember testing it... but I can't remember
including it or not.

 -- mnw

On Apr 21, 2016 8:57 AM, "dclow" [email protected] wrote:

Already found one bug in DCL-dev-1... the Markov solver uses linear
interpolation even when it is supposed to use cubic. I'm working on this
now.

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub
#32

from hark.

You did. I think I can get to the point where the new version returns the same answer as the old version, if told to use cubic splines. But if we aren't sure that's the correct answer I might not invest a huge amount of time in it... let's see how hard this turns out to be

from hark.

For what it's worth, the reason I didn't find this bug until now was I had set a_size to be much bigger (48), and so the (linear) interpolation results from the new version and the (cubic) results from the old version appeared identical. So that at least is an encouraging sign that the cubic results are right

from hark.

I'm pretty sure it's correct, from comparing the results of the linear and
cubic solvers of the same problem, examining behavior near the lower bound
of m (zero, in all cases I worked with), and comparing the results to the
TractableBufferStock model. We still need to fix the problem with gothic_h
in infinite horizon models, but that also applies to non-Markov models.

 -- mnw

On Thu, Apr 21, 2016 at 9:11 AM, dclow [email protected] wrote:

You did. I think I can get to the point where the new version returns the
same answer as the old version, if told to use cubic splines. But if we
aren't sure that's the correct answer I might not invest a huge amount of
time in it... let's see how hard this turns out to be

—
You are receiving this because you commented.
Reply to this email directly or view it on GitHub
#32 (comment)

from hark.

The Markov solver is now working correctly after MNW-dev-3 merge. Still requires OOP refactoring based on newest DCL-dev-1 commits, but this might be tricky.

from hark.