Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: RE: ttest or xtmelogit?


From   David Airey <[email protected]>
To   [email protected]
Subject   Re: st: RE: ttest or xtmelogit?
Date   Wed, 12 Mar 2008 10:37:17 -0500

On Mar 12, 2008, at 10:11 AM, Verkuilen, Jay wrote:

David Airey wrote:

I have a typical pilot data set.<
A few things:

(1) Why not try xtmelogit? Can't do any harm to see what it does.
Joseph Coveney posted code comparing xtlogit with ttests on summary statistics. The two results from his code relevant to my data were:

delta = 0.000 rho = 0.050

Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
delta | 300 .0037128 .3009899 -.9131391 .7421806
rho | 300 .0402976 .0260353 2.02e-10 .1294623

Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
pos_p_r | 300 .1 .3005013 0 1
pos_p_t | 300 .0466667 .2112763 0 1
pos_p_l | 300 .0466667 .2112763 0 1

delta = 1.000 rho = 0.050

Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
delta | 300 1.003721 .2986725 .0978231 1.812812
rho | 300 .0371925 .0277583 8.65e-10 .1519448

Variable | Obs Mean Std. Dev. Min Max
-------------+--------------------------------------------------------
pos_p_r | 300 .9233333 .2665064 0 1
pos_p_t | 300 .8566667 .3509979 0 1
pos_p_l | 300 .86 .3475668 0 1


In general I follow your advice, thinking that the right model for a small data set can't be a bad thing. Following Joseph's posts, I think the key to making informed choices is to be able to program simulations to see where each approach succeeds and fails. This is one of the primary advantages to Stata. Stata makes it easier to do this. Very easy for Joseph! Five times less easy for me. But I learn more every time someone posts code! Very fun.




(2) Resampling methods seem tailor-made for this kind of problem.
Stata's got really impressive support for clustering and stratification
for both jackknifing and bootstrapping.
That's a good thought.


(3) On a problem like this, I'd give serious consideration to a fully
Bayesian model.
I not facile with Bayesian methods in practice.

Thanks,

-Dave



Jay

*
*   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/
*
*   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 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index