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: conception confusion - "fixed effects" and time effect on data with time factor


From   Maarten Buis <maartenlbuis@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: conception confusion - "fixed effects" and time effect on data with time factor
Date   Fri, 21 Oct 2011 19:37:02 +0200

On Fri, Oct 21, 2011 at 6:58 PM, House Wang <jwanghouse@gmail.com> wrote:
> When you said "Time can have an effect on your outcome in
> the sense of aging or decay or it can be a proxy for everything that
> happened in a given year that in turn may have influenced your outcome
> variable." I think the time effect can be both. If so, should I have
> two separate models: one is for the sense of aging, the other is for a
> proxy for everything?
> May I put them together? What I did is like this: xttobit y1 x1 x2 x3
> year, i(year) ll(0)  ul(1).
> I write this model to see 1) random effects for year and  2) trend
> effect for year.
> The model works. The results show that there are no random effects for
> year, but some small but significant trend effects for the variable
> year. My interpretation about random effects is that there is no
> effect due to year as a proxy for everything that are confounding
> variable. If this is correct, I will not add year as a categorical
> variable. And I will go back to normal tobit model  and add year as a
> trend variable.
> But I am still not sure whether this model make sense.

You are right to be skeptical, so am I. Your model identifies the
difference between "trend time" and "proxy time" using the assumption
that the trend is linear, which is almost never true. You really need
to decide what it is that you want to measure, and and think of a
model that will get you that. From that, derive the role time plays in
your model (note model, not data or reality or ...). By focusing on
what you want to know, you can often trim your model to manageable
proportions. If you try to represent "reality" your model tends to
explode. In the end: models are supposed to be useful, not true.

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/


© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   Site index