vFunc and vPPfunc are sometimes Null Functions

[DominicWC]

Describe the bug
If PerfForesightConsumerType or IndShockConsumerType are solved, their solution contains a function 'vPPfunc' with the type HARK.utilities.NullFunc. This function can't be plotted. This is likely an issue in several other models.

For IndShockConsumerType specifically, vFunc is also a Null Function.

To Reproduce

from HARK.ConsumptionSaving.ConsIndShockModel import IndShockConsumerType
from HARK.ConsumptionSaving.ConsIndShockModel import PerfForesightConsumerType
agenta=IndShockConsumerType()
#agenta.cycles=0 #optional, these are here to show that whether the solution is a finite or infinite horizon doesn't matter
agenta.solve()
agentb=PerfForesightConsumerType()
#agentb.cycles=0 #optional, these are here to show that whether the solution is a finite or infinite horizon doesn't matter
agentb.solve()
print(agenta.solution[0].vFunc)
print(agenta.solution[0].vPPfunc)
print(agentb.solution[0].vPPfunc)

• OS: Windows 10
• Econ-ARK and Python: version 0.15.0, python 3.11 and 3.12

[alanlujan91]

vFunc requires vFuncBool = True and vPPfunc requires CubicBool = True

[DominicWC]

That fixes the IndShockConsumerType. The PerfForesightConsumerType is still returning a null function for vPPfunc though.

[mnwhite]

That's because there is no cubic spline interpolation for PerfForesight. The *true* consumption function really is piecewise linear (often just one line!), not smooth. So the marginal marginal value function is either perfectly represented by the (numeric) derivative of the marginal value function, *or* is actually a discontinuous function that we don't currently have a way to represent. Both LinearInterp and CubicInterp treat the function as continuous. We would need to extend those classes to allow versions with "proper" discontinuities. As is, discontinuities can be *approximated* by putting back-to-back gridpoints that are (say) 1e-12 apart. You *can* put two gridpoints right on top of each other, but it runs the risk of the function being evaluated *right* at that point, which will return an error without further intervention.

…

On Tue, Jun 11, 2024 at 10:57 AM DominicWC ***@***.***> wrote: That fixes the IndShockConsumerType. The PerfForesightConsumerType is still returning a null function for vPPfunc though. — Reply to this email directly, view it on GitHub <#1452 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADKRAFNOKIIWDBHVB22CDXDZG4F4VAVCNFSM6AAAAABJC3LTL2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNRQHE3TSNJYGQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[alanlujan91]

@mnwhite should it throw an AttributeError then? "You tried to set an attribute that is not applicable to this setting."

[mnwhite]

*Maybe*. I'm mostly ok with users doing slightly risky things like that.

…

On Tue, Jun 11, 2024 at 11:08 AM Alan Lujan ***@***.***> wrote: @mnwhite <https://github.com/mnwhite> should it throw an AttributeError then? "You tried to set an attribute that is not applicable to this setting." — Reply to this email directly, view it on GitHub <#1452 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADKRAFJPG2PMP6OL346H4OLZG4HGFAVCNFSM6AAAAABJC3LTL2VHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNRRGAYDKMJYG4> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[DominicWC]

It seems like this is working as intended. I'll stick something about this in the documentation.