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

Re: st: mixed effect model and autocorrelation

Subject   Re: st: mixed effect model and autocorrelation
Date   Sat, 13 Oct 2007 12:00:33 +0200

Panel commands in Stata (note: only "S" capitalized!) usually accept unbalanced panels as input. 
-glamm- (remember the dashes!), which you can download from ssc (by typing: -ssc install gllamm-), allow for the option cluster, which at least partially corrects for autocorrelation. Here you can find an useful intro on both the command and mixed models
But: are you sure that you need a multilevel mixed-effects command? -xtmixed- would be very useful e.g. in trying to predict a region level dependendent variable using region- and country-level independent variables, where regions are nested within countries. If you need something else, other commands may turn to be more useful and/or may allow for more options. An Kit provided useful suggestions for these cases.

At 02.33 12/10/2007 -0400, [ISO-8859-1] Ola Sj÷berg wrote:
>Hi everyone,
>This is the first time I try to post a question on the Statalist, and I 
>hope someone out there can help me to understand my perhaps somewhat 
>naive question.
>I have an unbalanced dataset, covering 15 countries over 20 years. As 
>usually with time-series data, there is significant autocorrelation that 
>I somehow need to take into account. One option for analyzing this data 
>would be what I would call "pooled time series regression", or the xtgls 
>option in STATA. These models would allow me to take into account the 
>autocorrelation. Another option would be a multilevel mixed-effects 
>model (xtmixed), which I understand have some nice features: it can 
>accommodate unbalanced data, and it will estimate variance components at 
>different levels (over time and between countries).
>Now to my question: How should I think about, and understand, the 
>autocorrelation in the mixed effects model? With my (somewhat limited) 
>experience with time-series data, I need to somehow take the 
>autocorrelation into account. How is this done in a mixed-effect model? 
>Since years are nested within countries, will the estimated standard 
>errors be valid?
>I hope someone can just give some hints on how to thick about this, or 
>perhaps give me some suggestions on readings.

*   For searches and help try:

© Copyright 1996–2015 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index