Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

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

RE: st: Fixed - effects

From   Maarten buis <>
To   stata list <>
Subject   RE: st: Fixed - effects
Date   Sun, 1 Aug 2010 10:47:30 +0000 (GMT)

--- Ignacio Pardo wrote:
> I read that it's not possible to generalize results to de population
> when using a Fixed Effects model (-xtlogit, fe-  or -xtreg, fe)

I am guessing that you are confusing the term "population average" with
the concept of being able to generalize to the population. Both models 
can generalize to the population (assuming that your data was collected
using an appropriate sampling scheme), but they differ with respect
to the definition of the unit of analysis. 

Terminology is alsways the main stumbling block when talking about 
this type of analysis, so lets use a concrete example: Assume we
have people nested in hospitals. A fixed effects model would look
at the people, while the population average model looks at the 
average response per hospital. The former is useful when we want to
look at how good a certain medicine works, while the latter is 
useful when we want to plan how many hospital beds, anti-viral 
medicine, and body bags a certain hospital needs to have in stock
for the next bad flu pandemic. So the choice between -xtlogit, fe-
and -xtlogit, pa- (or -xtgee-) depends on what the unit of analysis
is in your question, not on whether or not you can generalize to
the population, as both can do the latter.

A useful and readable discussion of this can be found in Chapter 13 
of Fitzmaurice et al. 2004. However, in that chapter they compare
marginal and mixed effects models. Marginal models are just a 
different name for population average models, mixed models are 
random effects models, but on this specific issue they are similar
to fixed effects models.

Hope this helps,

Fitzmaurice, G.M., Laird, N.M., and Ware, J.H. (2004) "Applied 
Longitudinal Analysis" Hoboken, NJ: Wiley-Interscience.

Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen


*   For searches and help try:

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