# deseasonalizing multiple series (more than 200 variables)

I'm trying to produce deseasonalization for multiple series using x-12 ARIMA (as an alternative, if you can manage, you also could provide an idea with other methods, such as x-13 ARIMA). The thing is that it does not seem viable to produce a thorough analysis of each variable and I would like to know if there is a standard practice (a standard model, I mean), less error prone, for doing such a type of deseasonalization. I've regressed the variables on seasonal dummies and saved the residuals as an alternative, but I would like to know if I could manage a deseasonalization with a more state-of-the-art approach.

Also, if you could hint a way (a code in R, Stata, Eviews8 etc.) of doing this analysis for producing multiple deseasonalized variables, I would be highly thankful.

• STL decomposition could be an alternative (function stl in R). – Richard Hardy Mar 26 '15 at 20:00
• Nothing comes close to unobserved components model, as this is the only decomposition method that i know of can seperarate trend and cycle, in addition it also lets you determine stochastic or deterministic trend/seasonality/cycle also identify outliers simultaneously. Not mention its ability to detect multiple seasonality. – forecaster Mar 27 '15 at 0:18

Default options are available and can be suitable when working with a large number of series. For example, if the parameter RSA=3 is used in TRAMO-SEATS (excerpt from the documentation): The program tests for the log/level specification, interpolates missing observations (if any), and performs automatic model identification and outlier detection. [...] The full output file is produced. X-12-ARIMA can be run with default options as well. In both cases, a text file can be created in order to choose the default procedure or other parameters of the procedure.