This project is read-only.

Filter Updating when missing data

Topics: Kalman Filtering
Apr 29, 2014 at 10:46 PM
Hi folks, let's suppose we have the following model:
R_t        = Beta1 * Infl_t + Beta2 * Ygap_t + e_t
Infl_t     = Beta3 * Infl_t-1 + e_t
Ygap_t     = Beta4 * Ygap_t-1 + e_t
I would like to know what/if parameters are updated when we have a missing value for R_t. I assume that Beta3 and Beta4 would be updated because those equations are not affected by the missing value, but Beta1 and Beta2 would not update. Or is it that none of the parameters would update?


May 1, 2014 at 1:43 PM
Just to make sure I understand.

You first run the estimate(...) command on a full set of observations for the three variables from startdate to enddate. Then, you add an extra observation on Infl and Ygap for period enddate+1, and run estimate(...) again from startdate to enddate+1.

I also assume that the three shocks all different althouh you use e_t to denote each of them.

Is this what you mean?

In that case, Beta3 and Beta4 should indeed be updated -- unless the new observations, by sheer luck, comply exactly with the model estimated previously, i.e. if Infl_t = Beta3Infl_t-1 and Ygap_t = Beta4Ygap_t-1 exactly for Beta3 and Beta4 estimated in the first run of estimate(...).

Marked as answer by jaromirbenes on 5/8/2014 at 9:22 AM
May 1, 2014 at 3:29 PM
That was exactly what I meant. Thanks!