Nobody has yet mentioned that we cycled around this
topic only a matter of weeks ago, in late November
and early December. Al Feiveson mentioned his -tryem-
Paul Millar his -bic- and I posted a new program
-tuples- which is discussed at
http://www.stata.com/statalist/archive/2006-12/msg00060.html
If -allpossible- (which I wrote) does what you want, that is fine,
but the approach allowed by -tuples- is more flexible. I
have not tried -tryem- and I have no prior probability
over -bic-, but their authors can discuss them.
Anyway, both -allpossible- and -tuples- are based on
precise the same idea as that mentioned by Michael.
Nick
[email protected]
Friedrich Huebler
-----------------
-allpossible- may be of interest to you.
. ssc d allpossible
SR Millis
---------
When may want to consider using Bayesian model
averaging. It is easily implemented in the free R
package BMA: it carries out BMA for linear regression,
generalized linear models, and survival analysis.
Available at the CRAN website:
http://cran.r-project.org/
Michael Verhofen
----------------
I have a regression with n different variables.
I would like to try all possible regression combinations.
For example (n=2):
regress yy1
regress yy1 xx1
regress yy1 xx2
regress yy1 xx1 xx2
I would like to try all possible regression combinations in a
convenient way, such that I can
enter the dependent variable and the independent variables.
Then, the program should
(1) find all possible 2^n model combinations,
(2) run each regression, and
(3) store statistics (such as _b, _se, R2).
A convenient way to find all model combinations might be to iterate
from 0 to 2^n-1 and convert the index variable to a binary number.
e.g. n=2:
model=0: 0 0 ==> regress yy1
model=1: 0 1 ==> regress yy1 xx1
model=2: 1 0 ==> regress yy1 xx2
model=3: 1 1 ==> regress yy1 xx1 xx2
*
* 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/
