energywise
energywise copied to clipboard
Initial values
Determine an inteligent m0 and C0 for the dlm model.
And/or assess the use of somewhat uninformative priors (e.g. normals with a lot of variance). The example code works like this, with state prior covariances ~1e7.
As well, given the context, it seems reasonable to make some assumptions about the possible ranges of these values.
It had also been suggested to use the series to chose a prior and then use it again?