Local linear trend model

I’d suspect you can either reparameterize the model or it’s the incorrect model for the data. There are a number of state space models to choose from here: GitHub - sinhrks/stan-statespace: Stan models for state space time series