Question:
In regression based approach for time series forecasting, I want to capture both exponential trend and multiplicative seasonality.
As per my understanding from the book and notes, I understand that we can use lambda = 0 in tslm() to capture exponential trend.
However, to capture multiplicative seasonality, do we need to take log(y) (i.e. log of dependent variable) also along with lambda = 0 in tslm().
Attached the book that I referred.
No, log(y) means including the argument lambda = 0 in the tslm function, so you have to do that once. To estimate exponential trend with multiplicative seasonality, very simply you can run the following code in R (Assuming your ts data is train.ts):
> train.lm.trend.season summary(train.lm.trend.season)
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression