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

[Maarten Buis <maartenlbuis@gmail.com>]

On Tue, Oct 18, 2011 at 7:58 PM, House Wang wrote:
> In this study, I am interested in the random effects of the year
> variable, which means the errors of D.V. due to year.
> To measure the random effects of year variable, the correct Stata
> command is to add i.year to the model, right?

There is some inconsistent terminology in your question. I guess you
want to do a regular regression (-regress-) and you want to know how
much of the variance in your dependent/explained/outcome/y variable is
explained by time.

The correct parametrization of time depends on how you think time
"works" in your problem. 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. In the former case you'd probably want some kind
(non-linear) trend, while in the latter case you probably want to add
it as a categorical variable like you suggested. However, you would
need to think really hard whether the variables you are by proxy
controlling for are in fact intervening or confounding variables. In
the former case you'll bias your results when you include time and in
the latter case you'll bias your results when you don't include time.
The problem is that probably some of the variables are intervening and
others confounding. I that case the best, but probably unrealistic,
solution is to directly measure the variables you want to control for
rather than rely on a proxy.

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/