Kalman state variables as observed variables

[redalert3]

On eviews, can we use "non-hidden" observed variables as state variables?

i.e. if we assume that household consumption is influenced by income and inflation, could we use income and inflation data as state variables?

I tried Kalman filter with sv1+sv2(i.e. random variables) and income+inflation, and found that the estimation results were the same.

@ename resid1
@ename resid2
@ename resid3

@evar var(resid1) = exp(c(1))
@evar var(resid2) = exp(c(2))
@evar var(resid3) = exp(c(3))

@signal consumption=income+inflation+resid1
@state income=income(-1)+resid2
@state inflation=inflation(-1)+resid3

Additionally, where can I check the spreadsheet for the Kalman filter outcomes?

Thanks much.

[trubador]

On eviews, can we use "non-hidden" observed variables as state variables?

No. Because, there is no point in doing so.

i.e. if we assume that household consumption is influenced by income and inflation, could we use income and inflation data as state variables?

No, they are already observed. You can however, filter the observed series and then use the latent part:

Code: Select all

... @signal consumption = core + base + ... @signal inflation = core + ... @signal income = base + ... @state core = core(-1) + ... @state base = base(-1) + ... ...

I should mention that, this is an ill-defined model and may not even produce meaningul/feasible results.

I tried Kalman filter with sv1+sv2(i.e. random variables) and income+inflation, and found that the estimation results were the same.

That is because, EViews believes what you say is true and treats income and inflation variables as unobserved. It does not double check whether they already exist in the workfile. Maybe it should...

Additionally, where can I check the spreadsheet for the Kalman filter outcomes?

Search the manual for state space data members.