[R] Three questions about DSE

[Rebecca Sela]

I have been using the dse1 and dse2 packages to estimate a model in which the underlying state is an ARMA(2,2) and the observed variables are equal to the state plus noise.  I am describing this model using a state space model.

First, in estimation, is there a way to restrict two of the estimated coefficients to be equal to each other?  In order to desribe an ARMA(2,2) model using a state space model, I must restrict the innovation variance to have its two non-zero elements equal to each other, and I haven't figured out how to do that.

Second, I am using the smoother to extract the signal using the code in the attached file.  The estimated value of the signal in the last period is 0, but all the estimated values are right.  Is there something I am missing here?

Finally, sometimes the smoother fails, but the filtered version of the signal is fine.  (That will not be the case with the attached code.)  Is there a known reason why this sometimes occurs?

Thank you in advance for all of your help!

Rebecca

--
Rebecca Sela
Doctoral Candidate
Statistics Group/IOMS
Stern School of Business