Once the series has been decomposed into the seasonal, trend, and remainder components with stl is there a 'predict' like function that I can use this data to forecast the next 'n' values? I didn't see one in the stl documentation. So if such a function does not exist I was wondering if I could take a straw poll on what methods are used for a time-series prediction with seasonal components? The problem with methods like 'ar' etc. is that in my case the "seasonal" components are a long "distance" from the beginning. I have daily data for a year and the "seasonal" components may only be during a few weeks of each year. Thank you for your help and suggestions. Kevin