Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Re: two-tailed test


From   Kit Baum <baum@bc.edu>
To   statalist@hsphsun2.harvard.edu
Subject   st: Re: two-tailed test
Date   Sat, 23 Oct 2004 09:02:30 -0400

On Oct 23, 2004, at 2:33 AM, Eric wrote:

(since you shouldn't really do two-tailed tests if you
have a theory and if you don't have a theory, you shouldn't be doing at
least social science--I'll let the hard sciences speak for themselves).

I don't know if Eric considers economics a social (antisocial?) or hard science, but I find this statement bordering on the ludicrous. Imagine a model

y = b0 + b1 x + b2 x^2 + e

From the standpoint of economic theory, there could be either a concave or convex relationship between y and x (that is, the second derivative is not signed by theory, depending on the shape of individuals' preferences). The data might be able to tell us that one of those cases exist, or they might be not sufficiently informative to reject the null b2 = 0. There is nothing wrong with a two-tailed test.

The ability to 'get one tailed tests automatically': is Stata automatically supposed to know which tail should be considered, or should it perform like 'ttest' and give you both? That IMHO would be more confusing for some people than their recalling the rule: if the test stat is on the wrong side of the inequality, don't reject; if it is on the right side, divide the pval by 2.

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.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/




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