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

Re: st: Issue with xtivreg and unbalanced panel

Subject   Re: st: Issue with xtivreg and unbalanced panel
Date   Thu, 19 Aug 2004 10:48:38 +0200

Good morning,

>We estimated a panel data model using the two following procedures:
>xtivreg y x1 x2 ( x3 = l.x3 )
>by panelvar: generate x3lag=x3[_n-1]
>xtivreg y x1 x2 ( x3 = x3lag )
>The difference between (1) and (2) is that in (2) we generate a new
>series for the lag of x3. We expected (1) and (2) to always produce
>same results, but that's the case only when the panel is balanced.

I suspect the answer is quite simple.

x3[_n-1] is conceputally different from l.x3.

x3[_n-1] refers to the previous value of variable x3 *in your data*. no
more no less. l.x3 refers to the value of x3 *at time t-1*.

If you have panel data, sorted by group id and time, and use

>by panelvar: generate x3lag=x3[_n-1]

x3lag will contain the previous observation of x3.

With unbalanced panel data, and if you have "holes" in the time-series of
observations for some groups (use -xtdes- to check that), l.x3 will only be
equal to x3lag if "year[_n-1] == year - 1" (here I assume year is your time
variable), ie. if data in the previous year is not missing (in which case
l.x3 is missing).

For the same reason, l.x3 will not be identified as often as x3lag, so you
must expect to have less observations in your regressions.

>When the panel is unabalanced, it seems that (1) drops entire
>while (2) doesn't. (2) reports much higher numbers for groups and
>observations than (1). In addition, we can find a number n such that
>following command (which explictly excludes several groups) gives
>same results as (1):
>xtivreg y x1 x2 ( x3 = l.x3 ) if freq>=n

What you describe is consistent with the above explanation, but it is hard
to ascertain that without knowing the pattern of missingness/holes in your
specific data. That's somethying you need to check by yourself (again
-xtdes- will help there).

Hope this helps.

Best wishes


Dr. Philippe Van Kerm
IRISS - Integrated Research Infrastructure in the
Socio-Economic Sciences
BP48, L-4501 Differdange, Luxembourg
>> <<
>> <<

This email and any files transmitted with it are confidential and
intended solely for the use of the individual or entity to whom they
are addressed. If you have received this email in error please notify
the system manager.

This footnote also confirms that this email message has been swept for the presence of computer viruses.


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index