GMM in EVIEWS

[Emmanuel Cabral]

Hi Guys!

I am trying to get the parameters of interest models using GMM. For example, for the CIR model, I used the method described by Chan et. al. in a 1992 paper (see attachment). I implemented the GMM estimation on actual daily 1-mo T-bills from 2002 to 2009. I am getting values for alpha, beta and gamma but with very low t-statistics. The p-values are very high. What could this mean?

Regards,
Emmanuel

 

[Emmanuel Cabral]

Sorry, I meant sigma, instead of gamma.

 

[HyperVolatility]

This could be caused by heteroskedasticity. In Eviews when you calculate parameters using GMM you can weigh the matrix by ticking the Cross Section box(White covariance matrix).
The software will estimate parameters weighting them according to the variance in the time series and this should lower their p-values.
However, should not that be the case you could try to smooth the series perhaps eliminating extreme values that make the variance go crazy.
hope this helps

www.hypervolatility.com

 

[Emmanuel Cabral]

Thanks! Will try that