How to impose lower and upper bound on a model parameter in unobserved components?

[Roland K]

When adding a cycle with unknown frequency to the observation equations, I can specify a search interval for the unknown period. Such an interval is a constraint when estimating the model parameters using ML. How can I do the same for other model parameters, e.g., sigma2.irregular?

[Chad Fulton]

As described in the Github issue (https://github.com/statsmodels/statsmodels/issues/9160), you can do this by subclassing the `UnobservedComponents` class and then using a parameter transformation to ensure the parameter stays within a set of bounds.

On Wed, Feb 21, 2024 at 1:20 PM Roland K <rkle...@gmail.com> wrote:

When adding a cycle with unknown frequency to the observation equations, I can specify a search interval for the unknown period. Such an interval is a constraint when estimating the model parameters using ML. How can I do the same for other model parameters, e.g., sigma2.irregular?

--
You received this message because you are subscribed to the Google Groups "pystatsmodels" group.
To unsubscribe from this group and stop receiving emails from it, send an email to pystatsmodel...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/pystatsmodels/8fe6e3ee-3674-450c-a34b-d939d54a527dn%40googlegroups.com.