Xiangping JIA:
As you correctly noted the p-value and the underlying test statistic
can be converted into one another, so they convey the same information.
I like the standard error since with some simple rules of thumb I can
see whether the test statistic is significant, and a confidence
interval. Which one of these is reported differs a lot between
disciplines, some just report one, two or three stars. On the one hand
I don't like it since you loose information that way, on the other hand
people tend to put too much confidence into p-values/standard
errors/test statistics (for instance every once in a while there is a
question on statalist where someone wants to know a p-value upto 16
digits) P-values/standard errors are themselves estimates and are thus
uncertain, certainly not up to 16 digits (you could for instance
bootstrap a p-value and get a confidence interval around it) and are
based on a model. So I am not so sure how much information is lost by
just "gazing at the stars". I believe there was quite a discussion in
psychology to just report confidence intervals. Which is yet another
way to display the degree of confidence in your results.
Anyhow, back to the question: p-values and their underlying test
statistic (and the standard error) confey the same information, so
there is no real way to choose between them. I like standard errros
since it makes it easier for me to get all three quantities (but it
being easier for me may just be because I am used to it). As to which
is conventional depends big time on the discipline. So if you made a
bet, than this answer won't help you choose a winner, sorry.
HTH,
Maarten
--- Jia Xiangping <[email protected]> wrote:
> When I draft the econometric report, I come across a font problem of
> reporting econometric results. As in most of journals, the
> coefficient and its corresponding significance should be reported,
> together with other indicators. While my colleagues are using t value
> or z value, I prefer to p>|t| because it is more informative and
> readers are informed at which level parameters can not be rejected.
> Though both of these types can be converted between each other, which
> one is better or more standard? We discussed this at lunch break and
> no agreement yet.
-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology
Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands
visiting adress:
Buitenveldertselaan 3 (Metropolitan), room Z434
+31 20 5986715
http://home.fsw.vu.nl/m.buis/
-----------------------------------------
___________________________________________________________
Now you can scan emails quickly with a reading pane. Get the new Yahoo! Mail. http://uk.docs.yahoo.com/nowyoucan.html
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
