[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: panel with cross-sectional dependence nd endogeneity

From   "Ivan Etzo" <[email protected]>
To   <[email protected]>
Subject   Re: st: panel with cross-sectional dependence nd endogeneity
Date   Thu, 20 Sep 2007 18:50:18 +0200 (ora legale Europa occidentale)

Thank you for your help Austin. As I pointed out before I'm new in panel methods...
-------Original Message-------
Date: 09/20/07 17:05:58
Subject: Re: st: panel with cross-sectional dependence nd endogeneity
Ivan Etzo <[email protected]>:
Forgot the 2 in -xtivreg2- (-ssc inst xtivreg2- to install).
-xtivreg- is official Stata, and does things that -xtivreg2- does not
(RE), but -xtivreg2- produces more diagnostics and has more options
for "fixing" the SEs.
On 9/20/07, Austin Nichols <[email protected]> wrote:
> Ivan Etzo <[email protected]>:
> It sounds like you should read a shelfful of books on panel methods,
> though you won't find the magic bullet you want.  I would recommend
> you start with -xtivreg- from SSC and use the cluster(id) option where
> id identifies panels, then read BSS2007 on some other options:
> ps. plain text only emails on Statalist, please!
> On 9/20/07, Ivan Etzo <[email protected]> wrote:
> > my panel (N=2660 ; T=7) presents either heteroskedasticity (lrtest and xttest3) endogeneity and cross-sectional dependence (xtcsd test). Moreover there is also serial correlation (xtserial test).
> > Thank to statalist I found useful commands to correct my analysis but I haven't found a way to cope with all of them together yet. The command xtscc for ex. is useful for the cross-sectional dependence (nd maybe also heteroskedasticity..?) but not in the case of endogenous covariates. For the endogeneity problem, I tried different commands like ivreg2 and xthtaylor, the last one give nice results in terms of expected coeff signs and their significance, but they don't provide correct SE for cross-sect dependence. Xtgls seems good for heteroskedasticity (and serial correlation) but not for endogeneity.
> > I can say that I'm new with empirical analysis and panel data so every suggestion or comment would be pretty appreciated.
*   For searches and help try:

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