Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Maarten Buis <maartenlbuis@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: First-Differece with or without a constant? |
Date | Mon, 7 May 2012 15:52:51 +0200 |
---On Mon, May 7, 2012 at 1:47 PM, Hawal Shamon wrote: > I like to estimate a First-Difference-Model on the basis of two waves. > > Some literature (e.g. Wooldridge 2008) recommends to estimate First Differenceusing the constant as follows: Such literature reference with only an author name and year are not appreciated on this list. Remember that this is an interdisciplinary list: references that are so "World Famous within your (sub-(sub-))discipline" that an author/year reference suffices, are likely completely unknown outside your micro-cosmos. This is clearly explained in the Statalist FAQ which you were asked to read before posting. > [Delta]yit = [alpha]0 + [Delta]x1it + . + [Delta]x2it + [Delta]eit , > > where [alpha]0 denotes the difference of the intercepts of y for both years which is nothing else than the change. A disadvantage occurs when any change in x ([Delta]xkit) does not vary between the units. E.g., having a panel dataset with employees over two subsequent years means that job experience is increasing for all of them over the two subsequent years by one year. In this case [Delta]x1it will be dropped due to collinearity. This is not a problem, just a mistake by the analyst: It makes no sense to add a "variable" that does not vary. The solution is not to leave out the constant, but to leave out the offending "non-varying-variable". > What do you think? Which model is "in general" the better one and why? In general all you need is to understand the argument you are making when using a given model and effectively communicate that with your audience. More specific advise like "always add a constant" tends to degenerate into a "cookbook-style" of statistics that generally does more harm than good, e.g. <http://www.stata.com/statalist/archive/2010-12/msg00614.html>. -- 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/