06-20-2017 12:00 PM
Thank you. I gather there is no way to alter the damping factor at all to delay the damping factor to later forecasted dates? It tends to converge rather quickly.
06-20-2017 02:39 PM
Not with proc ESM, it does not allow to specify the value of the parameters. The parameter values are optimized based on the data.
I would suggest looking also at UCM models (PROC UCM). UCM models are somewhat similar to EMS models, in that they decompose the series into components. They do not allow multiplicative models, but they are more flexible in terms of the components, allow the user of regressons, and you can specify the values of the parameters if you wish.