I'm trying to recreate a model from Stan that uses a quadratic exponential covariance form for a spatial model (Stan code here). Once the covariance matrix is built they take the cholesky decomposition of it.
When I try this in RTMB it complains about the matrix being complex. I know it's not but presume this is because RTMB uses complex numbers on the backend. Is there an easy way around this, or an alternative approach?
kaskr
commented
2 months ago
I'm trying to recreate a model from Stan that uses a quadratic exponential covariance form for a spatial model (Stan code here). Once the covariance matrix is built they take the cholesky decomposition of it.
When I try this in RTMB it complains about the matrix being complex. I know it's not but presume this is because RTMB uses complex numbers on the backend. Is there an easy way around this, or an alternative approach?