Thanks Allin - and thanks Sven for the earlier e-mail.
I am fully covered.
Alecos Papadopoulos PhD
Affiliate Researcher
Dpt of Economics, Athens University of Economics and Business
Foundation for Economic and Industrial Research (IOBE)
web:
alecospapadopoulos.wordpress.com/
ORCID:0000-0003-2441-4550
On 22/5/2024 23:14, Cottrell, Allin wrote:
> On Wed, May 22, 2024 at 3:36 PM Alecos Papadopoulos <papadopalex(a)aueb.gr>
wrote:
>> When we combine the Kalman Filter with the mle command, we use the
>> "kfilter" or the "ksmooth" function as a success/failure
check for KF,
>> and if we're good, we tell the mle to have a look at the series with the
>> likelihood contributions, and proceed with parameter estimation.
>>
>> We automate this by using a conditional statement on the mle line itself
>> (as shown e.g. in the user guide, p. 362).
>>
>> In no-KF mle, this conditional statement is often used to impose
>> constraints on the parameters (like positivity of variances).
>>
>> I understand that simple non-negative constraints can be imposed in
>> other indirect ways, but I have to impose bounds on the estimated
>> parameters like "0<alpha<1".
>>
>> So can I somehow combine the naked mle "check" syntax with the
>> KF-success syntax into one, say
>>
>> <hansl>
>> mle logl ERR && CHECK ? NA : kb.llt
>> ....
>> ERR = kfilter(&kb)
>> scalar CHECK = (alpha>0) && (alpha<1)
>> params alpha
>> end mle
>> </hansl>
>>
>> ...or do I need something more elaborate? (maybe there is an issue with
>> the fact that good news for ERR is when ERR=0 while good news for CHECK
>> is when CHECK = 1, if I understand correctly?)
> Yes, that would be an issue in the setup as given above, but it could
> be circumvented by
>
> mle logl (ERR || CHECK == 0)? NA : kb.llt
>
> or, if you prefer, make the NA condition "(ERR || !CHECK)". However,
> if you want to limit a parameter to (0,1) a more standard approach
> would be to use a transformation that maps from (-inf, +inf) onto
> (0,1) -- for example cnorm().
>
> Allin Cottrell
> _______________________________________________
> Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it
> To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it
> Website:
https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/