The RATS Software Forum

Dear Doan:

I have a problem on transforming raw data. For example, I have a quarterly real GDP series, I want to use the UC model to decompose it into the trend and cycle. I know I must log and seasonal adjust it. The problem is the order. At generally I do first seasonal adjustment the origin series, then logarithm. Somebody suggests the opposite order. Pls tell me right answer.

Best Regard
Hardmann

First, note that there is a form of UC which includes a seasonal component (actually one of two types of seasonal components---an additive for seasonals which are primarily calendar-driven, or a harmonic for those that are temperature-driven).

The most common form of X11 adjustment nowadays is the log-additive which is very similar to what you by an additive adjustment on log data, but not identical. You would then (in practice) typically take the log of the log-additively adjusted data.