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: xtlogit and clustering

From   Jeph Herrin <>
Subject   Re: st: xtlogit and clustering
Date   Sun, 28 Nov 2010 11:40:52 -0500

With only 2 schools you will certainly run into trouble trying model
variance between schools.

More generally, you can estimate 3-level logit models with -xtmelogit-.


On 11/27/2010 9:29 AM, Shehzad Ali wrote:
Thanks, Maarten. That sounds like a good idea.

The option I was wondering may be:

gllamm y x1 x2, link(logit) fam(bin) i(school id)

Here obviously the id (the student id) is the panel variable and school is the
higher level cluster variable. I wonder if this would be a wise option or should
I stick to what you suggested? Also, following on from your advice, the school
variable was not statistically significant when introduced as a covariate. Is
this an indication that I shouldn't worry about the school level clustering?


----- Original Message ----
From: Maarten buis<>
Sent: Fri, 26 November, 2010 15:56:38
Subject: Re: st: xtlogit and clustering

--- On Fri, 26/11/10, Shehzad Ali wrote:
If you have clustered data from  two schools with repeated
observations for students at three time  points. Is it
possible to use the -xtlogit- model and  specify  that
clustering occurs at the level of students (the panel
data)  and  also at the level of schools (two schools)?

Why not just  include a dummy for school? Two clusters
seems to me way too few for anything  more sophisticated.

Hope this  helps,

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

*   For searches and help try:

*   For searches and help try:

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index