Bookmark and Share

Notice: On March 31, it was announced that Statalist is moving from an email list to a forum. The old list will shut down on April 23, and its replacement, is already up and running.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Fixed - effects

From   Clive Nicholas <>
Subject   Re: st: Fixed - effects
Date   Sat, 31 Jul 2010 23:03:14 +0100

Ignacio Pardo wrote:

> Today a short and easy stat question!

Nick Cox has stressed this before: please don't write this. It
irritates fellow Statalisters. Why? Well, if it really is an easy
question, why are you asking the list for the answer when you can
figure it out for yourself? Although it's the weekend, it's no
coincidence that my reply to your question is the only one you've had
in nearly nine hours.

> I' trying to create a F.E. model with an outcome as response and other
> outcomes as predictors...
> I read that it's not possible to generalize results to de population
> when using a Fixed Effects model (-xtlogit, fe-  or -xtreg, fe)

If it wasn't (meant to be) possible to generalize FE models from your
sample to the population, then why does -xtreg- have a -pa-
(population averaged) option? Do you what clearly haven't done, and
look at -help xtreg- closely.

Even if you concluded that the -pa- option to -xtreg- was unsuitable
for your model(s), there is still the bootstrapping option. For

set seed `=date("2010-07-31", "ymd")'
set obs 360
g id=_n
g group=ceil(uniform()*12)
tab group
expand 2
g y=uniform()*100
g x1=uniform()
g x2=uniform()*5
g x3=invnorm(uniform())
bysort id: gen time=_n
l id group y x1 x2 x3 time in 1/10
tsset id time
bs "xtreg y x1 x2 x3, i(id) fe robust cluster(group)" _b _se, size(360)

However, the usefulness of the bootstrapping approach depends upon
your units. If your units are, say, people or SMEs (small-to-medium
sized enterprises), then one could conceive that the population size
is large enough for it to be approaching infinity; therefore you could
bootstrap your data as if it were a stochastically-generated sample of
that population. But if your units are countries, using the
bootstrapping approach may be difficult to defend, since countries
don't emerge from a stochastic process, but rather historical and
socio-political ones.

So choose your method wisely depending on your data.

Clive Nicholas

[Please DO NOT mail me personally here, but at
<>. Please respond to contributions I make in
a list thread here. Thanks!]

"My colleagues in the social sciences talk a great deal about
methodology. I prefer to call it style." -- Freeman J. Dyson.

*   For searches and help try:

© Copyright 1996–2016 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index