[R] StructTS; measurement variance zero

Andreas Poller Andreas_Poller at gmx.de
Sun Sep 19 11:26:40 CEST 2004

Hello everybody,

I'm investigating several time series with StructTS, and for one series 
I get zero variance for the measurement errors (local level model).
I've read in Brian Dipleys article on time series in the R-Newsletter 
June/2002, that the airline passengers data set yields the same result.

Having the variance of epsilon as zero, I would expect the residuals (y 
- Z a) to be zero.  The substraction of data and fitted values for the 
states in fact yields zero, but the output of the KalmanRun for the 
residuals does not. So I don´t understand what´s going on.

Could someone help me with this problem?

Tanks a lot!


More information about the R-help mailing list