Statalist


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

RE: st: Negative LR test statistic ?


From   "Lachenbruch, Peter" <[email protected]>
To   "'[email protected]'" <[email protected]>
Subject   RE: st: Negative LR test statistic ?
Date   Tue, 22 Dec 2009 09:00:01 -0800

A general rule of thumb is that the number of observations should be about 10 times the number of predictor variables for a single linear regression - it's not absolute, but seems to hold fairly well.  Thus, with 14 equations, you would probably not want to have much more than 5 or 6 predictors.  

Happy holidays all,

Tony

Peter A. Lachenbruch
Department of Public Health
Oregon State University
Corvallis, OR 97330
Phone: 541-737-3832
FAX: 541-737-4001


-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf Of Maarten buis
Sent: Monday, December 21, 2009 3:08 PM
To: [email protected]
Subject: Re: st: Negative LR test statistic ?

> --- On Mon, 21/12/09, Ekrem Kalkan wrote:
> > I am estimating  a system of 14 equations, each with
> > nearly 40 variables. I have also  20 excluded instruments. What
> > do you mean by "empty"model?  If you mean the model without
> > explanatory variables, there will be only 14 constant term to
> > be estimated. Is it too large?

--- I answered: 
> I am afraid that this could very well be the case. Think of it 
> this way: you have only a bit more than 60 observations per
> equation. 60 is OK but not great for linear regression, as it 
> is known to be robust, well behaved, and stable, but your are 
> realy pushing your luck when using such small sample sizes for 
> anything more complicated. This is especially true for anything 
> involving instrumental variables, these models can easily eat 
> huge amounts of statistical power. 

Let me add to that: 40 covariates would be way too much with only
60 observations, even for a linear regression. What you could do 
to get a feel for how much your data can take, is to do a power
analysis as described here:
http://www.stata.com/support/faqs/stat/power.html  

Hope this helps,
Maarten

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

http://www.maartenbuis.nl
--------------------------


      

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index