[Arpita Dey]

I couldn't get your question.
Are you talking about CI of inflation after xtgls, panels( hetero)? If this was your question then 95% CI of the inflation coefficient -.0085278 .012962.

[Carlo Lazzaro]

Arpita:
yes, that was the question.
Therefore, you cannot rule out that the coefficient of your inflation variable is negative but its real, fixed and unknown value can well be 0 or something else in the population from which your sample was drawn.

[Arpita Dey]

well thanks. Can you tell me why the signs are different when I am using xtgls, panels(hetero) instead of xtgls,igls panel(hetero) ?

[Carlo Lazzaro]

Arpita:
as under both the coefficients do not reach statistical significance, their sign is irrelevant, as the do not have any bearing on the conditional mean of the regerssand.

[Arpita Dey]

well. Another query is incased of N>T,I should stick with xtgls[ dept var,] [indep var ] if there is heteroskedasticity right ?

[Carlo Lazzaro]

Arpita:
if N>T, you should go -xtreg- with cluster-robust standard errors in case of heteroskedasticity and/or autocorrelation.