/* README FILE FOR USE WITH FIXED1.PRG AND RAND1.PRG */


These programs are used to generate data and perform a Monte Carlo analysis of various estimators of the dynamic panel data model (readers should be familiar with the numerous variants - see Harris and Matyas, 1996, for a discussion).

As such they should run without any modifications.

They have the facility to misspecify the assumed true data generating process (Harris, Longmire and Matyas, 1996, 1998) although the additional parameters (eg rho) ARE NOT ESTIMATED.

The GMM-type estimators are numerically dependent on the sample size and data set used, ie it is possible (probable) that not all of the assumed othogonality conditions can be used, and this will vary with the sample size EVEN IF THE DGP IS UNCHANGED.

Ther are several variants of these programs available, but vary only with the sample size(s), # of GMM conditions used and in the way the dependent variable is generated.

For use on a real data set, they would have to be ammended such that: N and T are defined as a function of the data set; the various Y's, X's and Z's (the instruments) would have to defined in terms of the actual data and; the Monte Carlo loop removed.

The programs were only intended for personal use, and therefore have not been fully annotated and may contain elements of testing, checking etc.