Page 1 of 1
kalman filtering with missing data
Posted: Thu Sep 04, 2014 1:53 pm
by EviewsUser1
I am trying to use the Kalman filter to estimate a dynamic factor model on a panel of data that is unbalanced -- there are missing observations at the end of the sample for a few of the series. Does Eviews automatically handle the missing observations in the Kalman algorithm? How do I specify this model in Eviews?
Re: kalman filtering with missing data
Posted: Thu Sep 04, 2014 4:20 pm
by EViews Glenn
The Kalman filter allows for missing values in some or all of the signal variables, but not in the regressors in the state equations.
Note that EViews state space tools do not work in panel settings.
Re: kalman filtering with missing data
Posted: Mon Sep 15, 2014 7:43 am
by EviewsUser1
Can I specify the state equations in matrix form?
F(t)=AF(t-1)+Bu(t)
where F is a 10x1, A is 10x10, B is 10x2 and u(t) is 2x1
Re: kalman filtering with missing data
Posted: Mon Sep 15, 2014 9:54 am
by EViews Glenn
Unfortunately not. You'll have to specify each of the 10 state equations individually.