Am 29.12.2020 um 13:08 schrieb Brian Revell:
here's another thought for the GUI structural time series modelling
options. Currently the Basic Structural Model only has a seasonal
cyclical element. Were this to be flexible and allow a user defined
periodicity/frequency for annual series, then longer term cycles could
be incorporated into the BSM. Admittedly, a trig cycle can be added
into the exogenous variable group, but it will be deterministic, whereas
a stochastic cycle would be useful (as most of those I have looked at
are not completely regular).
Hi, please don't hijack a thread's topic by simply replying. Start a new
thread or reply to the previous adequate one.
Having said that, as a reminder I'm not the only author of the StrucTiSM
package, but it's OK to address me personally. In fact, we prepared an
update that should hopefully solve your recent forecasting problems, but
it is in the moderation process and not quite published yet. The next
priority is to allow exogenous variables in forecasting.
I don't think we're going to extend the BSM, because by construction it
stands for a certain set of features. But StrucTiSM also supports more
general setups, BSM is just a subset.