Statalist The Stata Listserver

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

st: RE: Observation number in fixed effect regression

From   "Steven Stillman" <>
To   <>
Subject   st: RE: Observation number in fixed effect regression
Date   Sat, 21 Oct 2006 15:26:57 +1300


xtreg is only using the 200 observation with multiple siblings for the
majority of the results when running a fixed effect model (the exceptions
are the overall constant and the between r-squared), but by convention lists
all valid observations in the sample size.  you can see this yourself by
restricting the xtreg to the sample with multiple siblings.  personally, i
do not like this convention and typically include the appropriate if
statement to exclude singleton observations.


-----Original Message-----
[]On Behalf Of Hui Wang
Sent: Friday, October 20, 2006 3:17 AM
Subject: st: Observation number in fixed effect regression

Hi all,

I want to run the following regression:

Y = pai0 + pai1*A+X*pai2+e

Each observation is a kid in certain households. 'A' is the variable I am
interested in; "X" is a vector of other controls. There are totally 1100
kids. Some of them are siblings in the same houshold. The number of kids who
have siblings in the dataset is about 200. If I run the following command in

xtreg Y A X , fe i(id)

where id is the houshold ID. I should expect the stata to explore the
variation within the households with multiple kids to estimate pai1. That
says, the observation number should be about 200. But in the results, stata
report that the number of observation is actually about 1100 (the whole

Why is that? Is there any way to correct this? I found that the same problem
does not happen to  logit regression (xtlogit).



Express yourself instantly with MSN Messenger! Download today it's FREE!

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2021 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index