12-02-2011 01:07 PM
I used the PDL model (with AR(1)) to deal with mutiple years data on the weekly basis. However, I have missing data in some weeks.
I understand the model will ignore the observation with missing value.
The problem is the AR(1) structure. When I plot the comaprison bvetween observation and predicted value, the predicted value seems have one step behind the OBS.
How can I deal with this probelm? Thanks!