EViews.com

Dears,

I've just computed the p-values of Sargan test for over-identifying restrictions in GMM Dynamic Panel Data Model.
I got the p-values by expressing the following
scalar pval=@chisq(x,y) where x is J-statistics and y is n. of instruments-n.of coefficients

However I have some doubts how to interpret the p-values since different packages have opposite ways of expressing this statistics.

Namely if my p-values are of the order of 0.00001 circa, shall I reject the null of valid over-identifying restrictions?

thanks

Hello, I have the same question. How can be explained this? Thank you.

While different packages have different ways of writing the statistics, the p-values should be the same. A small p-value means reject.

Hello,

My data has N<T, I know that DPD estimator is essentially efficient/unbiased for N>T. But in an attempt to replicate a similar study I prefer to use it.
Total panel obs : 231
I used DPD estimator and model ran fairly well.
Number of Coefficients estimated 3.
J-stat : 4.456718
Instrument rank : 7 with p_value 0,26 where Sargan test null is not rejected and I'm happy.
But my instructor commented that the degrees of freedom in Sargan chi square test (7-4 =3) is low and poses a problem.
But does not point out what is the problem. I couldnt find a clue in books, forums as to whether there is a problem or not.
Since I am aiming to finalize the project would very much appreciate your assistance. many thanks in advance.
BR
Ahmet