[R] experience with nvmax(max size of subsets to examine)

CZ cxzhang at ualr.edu
Fri Oct 8 23:38:11 CEST 2010


Hello,

I am using regsubsets function in leaps package.  And I did some experiments
with the parameter nvmax(  max size of subsets to examine) and got some
weird observations.  

What I did is:

I used around 100 or about 600 variables and called function
regsubsets(data, nvmax=1...), I got a matrix (summary$which) consists of
only 1-variable models, which is I assumed to get. 

However, when using 1000 or 2000 variables, and calling the same function
(still nvmax=1),  the matrix of summary$which consists of both 2-variable
models and 1-variable models.  I wouldn't expect any models with 2-variables
by setting nvmax=1.  Similarly, when setting nvmax=2 and using more than
1000 variables, I also got models with 3-variables.

It could be my understanding on nvmax is wrong or some other reason.  Does
anyone have some idea on this?  Thank you.

CZ


-- 
View this message in context: http://r.789695.n4.nabble.com/experience-with-nvmax-max-size-of-subsets-to-examine-tp2969099p2969099.html
Sent from the R help mailing list archive at Nabble.com.



More information about the R-help mailing list