What z-stats should be expected in State Space
Posted: Mon Apr 28, 2014 8:16 am
Hi all,
I estimate a state space model of a futures curve. It has many datapoints: thousands of daily data and eight measurement equations.
The model fits well, it seems reasonable for Eviews to report a high Log likelihood (30-40,000) given the big number of datapoints.
But the standard errors of the coefficients are excessively small, giving z-stats of 100 to 300. That seems wrong.
How does Eviews calculate standard errors? Is a correction required? What estimate for standard errors should be expected?
Thanks, James
I estimate a state space model of a futures curve. It has many datapoints: thousands of daily data and eight measurement equations.
The model fits well, it seems reasonable for Eviews to report a high Log likelihood (30-40,000) given the big number of datapoints.
But the standard errors of the coefficients are excessively small, giving z-stats of 100 to 300. That seems wrong.
How does Eviews calculate standard errors? Is a correction required? What estimate for standard errors should be expected?
Thanks, James