Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: panel data analysis

From   "M H" <[email protected]>
To   [email protected]
Subject   st: panel data analysis
Date   Thu, 15 Apr 2004 13:33:57 -0400

Dear Stata Users,

Please forgive the lengthy post, but I wanted state up front the procedures
that I have already attempted, and the end result.

I would like to model set of firms (i=1 to N, where i=1,2,3 �. 396)  over a
10 year period using quarterly data (t=1 to T, where t=1,2,3�..396).  Based
on this framework, the panel is balanced (I have the same number if time
periods for each firm with no missing values) and the total number of
observations NT = 15,840.

Procedure 1: After fitting a model to the above panel of firms  using xtreg,
fe xttest0 and xtreg, re xttest1, I find, that the disturbances exhibit
serial correlation, cross correlation and heteroscasdacity (consistent with
my priors).

Stumbling block to Procedure 1: the xtreg does not provide an option to
correct for these problems with the error terms.  This becomes particularly
problematic as the tests for fixed and random effects rely on the assumption
that the disturbances are �well-behaved�.

Procedure 2: To address the problems with the disturbances mentioned above,
I decided to use xtgls, since this procedure explicitly addresses serial
correlation, cross correlation and heterosckedasticity.

Stumbling blocks to Procedure 2: xtgls does not allow for fixed or random
effects. Once way around this problems is to simply create dummies for each
of the cross sectional units (e.g. firms in this case) and omitting the
intercept.  Typically this would work, except to address the problems with
the error terms, T>=N. Since my data don�t conform to this condition, I
can�t use this module.

Other Procedures attempted:  I�ve tried a number of other procedures (areg
with the robust errors options and using dummies for the cross sectional
units) but of course this only addresses heterosckedasticity but not cross
correlation or serial correlation.

I�m not sure where to go from here and feel a bit like I�m chasing my tail.
Any suggestions would be greatly appreciated.

Limited-time offer: Fast, reliable MSN 9 Dial-up Internet access FREE for 2

*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index