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, statalist.org is already up and running.


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

Re: st: state trends


From   sara borelli <saraborelli77@yahoo.it>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: state trends
Date   Tue, 19 Oct 2010 19:09:16 +0100 (BST)

Hi Marteen,

thank you very much for your reply; I apologyze for asking further.

I started by analyzing each state. So what hapens is the following

when using the orginal year variable I have that for each state there is a perfect correlation between 
state dummy, trendstate, trandsquare_state  when these are constructed by using the orginal year variable

when I look at the correlation between the same variables constructed using the rescaled year variable then the correlation is very high but not perfect.

So this is why stata is dropping the state dummy in the first case.
However I do not understand why there is perfect correlation in the first case but not in the second. Sorry I know this may be a simple question but I am bit confused and I am trying to get this right

thanks 
Sara





--- Mar 19/10/10, Maarten buis <maartenbuis@yahoo.co.uk> ha scritto:

> Da: Maarten buis <maartenbuis@yahoo.co.uk>
> Oggetto: Re: st: state trends
> A: statalist@hsphsun2.harvard.edu
> Data: Martedì 19 ottobre 2010, 18:02
> --- On Tue, 19/10/10, sara borelli
> wrote:
> > Is that correct to say that in any case the
> coefficient of
> > my variables of interest will always be the same
> regardless
> > of whether I use year=1990,1991  or its scaled
> > counterpart? Just the coefficient on state fixed
> effects
> > change and the units in which the trend is measured
> should
> > not matter...is that correct?
> 
> Depends, as long as there are no interactions with time the
> 
> unit should not matter, though I would opt for some
> reasonable
> scale and origin for your research, which is often not
> years 
> since the year 0.
>  
> > I am not sure I understand the reply to the second
> question
> > (quadratic trend). If the quadratic trend based on
> orginal
> > year variable causes a drop in state fixed
> effects  why
> > it does not happen the same when I use the quadratic
> trend
> > based on the re-scaled variable? Would be 
> correct at
> > this point to use the quadratic trend based only on
> the
> > re-scaled year variable?
> 
> I can think of two reasons: First, you added your quadratic
> terms
> as a variable list var1-vark, and there could be other
> variables
> in that list than you intend. That was my first hunch. 
> Second, you may have ended up with precision and
> multicolinearity
> problems by squaring years that range between 1990-....,
> while 
> you to a large extend reduce those by sqaring variable
> ranging 
> 1, ...
>  
> As I said in my previous post: The best strategy for
> building 
> such models, is not to do everthing at once, but instead to
> break
> the problem up into manageable parts. In your case I would
> solve
> the trend issue for one state, and once you got that
> working, try 
> to extend this to your larger model by adding the
> interactions.
> 
> 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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index