Time series model

[Jacquie]

Actually there was a previous post on a similar topic, sorry I had missed that before:

Trend-cycle time series - convergence or mixing issue?

I’m trying to roughly replicate Harvey, Trimbur & van Dijk (2005). See pages 3-4 for their model. Right now I’m only concerned with 1st order cycles. As it stands right now, autocorrelations are really high - when I look at trace plots, several parameters (especially lambda and rho) look like they’ve just not really moving. Combining this with large Rhat values makes me think that it just hasn’t converged by the time it gets to sampling. What can I do to improve convergence? It seems like I h…

which seems to tell me that I should vectorise and use the ‘matt trick’ … trying this now