[Gretl-users] Re: The function works fine, but the package doesn't!
On Mon, Mar 10, 2025 at 10:10 AM Cottrell, Allin <cottrell(a)wfu.edu> wrote:
On Mon, Mar 10, 2025 at 7:22 AM Paolo Chirico <paolo.chirico(a)uniupo.it> wrote:
>
> I have prepared a function to estimate stochastic volatility models with three
different QML procedures (which I attach with a datafile).
> The first procedure is the classic QML, while the other two use time-varying state
space models, so they use update functions.
> The script runs fine with all the procedures, Jack has already seen it.
> However, after creating the package, only the first procedure works well, while with
the other two I get the message:
>
> This function needs a dataset in place
> The formula 'ERR = kfilter(&SVM)'
> produced an error on execution
> *** error in function svm_qm2, line 133
Paolo,
Your package should work if you set its "data-requirement" to
"no-data-ok" (either via spec file or the packaging GUI).
I have to say that at this point I don't understand why that's necessary, but
I'll try investigating further.
OK, now I see what was happening. The error was, as you thought,
related to the TV* updater functions. But there was nothing wrong with
your hansl code, rather the built-in kfilter() function was performing
a misguided check on the updaters. This is now fixed in git, and the
fix will be in gretl 2025a, due out soon. In the meantime a
not-very-nice workaround is as I said: pretend that your package
doesn't need a dataset in place.
Allin