Panel Estimation

[jiya]

I am estimating a fixed effect model in Eviews 7. When i estimate the model without correcting heterscadasticity I get the result with AIC. But the moment I choose GLS weights : Cross-section Sur and coef covarience method: white cross-section, I dont get AIC/SIC value in the dispaly. Why does this happen?...how can i choose optimal lag length of my interest explanatory variable without thsee criterion?

[EViews Glenn]

You need to iterate the weights to convergence.

[jiya]

How can i do that?...would u kindly elaborate a little more, please.

[EViews Glenn]

On the options page, there is a setting for iteration.

You want one of the "Iterate weights and coefs" settings.

[jiya]

yes there is an option and by default eviews 7 set this iterate coefficients to converge (under "update weights once, then..."). However, there are other options "Iterate weights and coefficients" but I dont know when to choose this option. I hv read the Eviews mannual but I dont find any explanation regarding situations when one can choose a specific options.
Actually, my problem is regarding "optimal lag length" selection and wanna knw hw can i do that in panel regression.

[jiya]

when i choose "simultaneous updating" Convergence is achieved after different weight iteration for different lags....next what?

[jiya]

sir kindly response: when should i choose "simultaneous updating", "sequential updating" , "iterate coef to converge" or "update coefs once".....please give a reply.

[EViews Glenn]

The original question was how to get the IC... Simultaneous or sequential iteration to convergence will tell EViews to produce those statistics.

[jiya]

yes...they do...thank you so much.

[jiya]

Although this is not related to my original question still I want kw: when should i choose "simultaneous updating", "sequential updating" , "iterate coef to converge" or "update coefs once".....please give a reply.

[EViews Glenn]

That's really up to you.

The first two do updating of the weighting matrices, but if you have a nonlinear specification you may get different results. The latter two only compute one weighting matrix iteration and differ in whether the nonlinear specification is iterated to convergence. For a linear specification the first two should be the same and the last two should be the same so the only question is whether you want to iterate the weights to convergence or not.

[jiya]

that means one can choose a particular one depending on his/her own choice...that is; one that best fits the model (sine Adj R2 is sensitive to these alternatives)

[runawayyy]

I am using Eviews 6 to estimate a dynamic panel model using unbalanced panel data and GMM/DPD method.
In the result window, I see neither R2 nor Adj.R2 !!
Why is that? and what should I do to know R2 of the model?
Thank you.

[ceshiwuhao]

You need to iterate the weights to convergence.

[runawayyy]

Can you be more specific?
What exactly should I do?