[R-pkgs] New version of Subselect package

[Jorge Cadima]

Version 0.9-99 of package "subselect" is now on CRAN. 

The subselect package has functions that search for k-variable subsets
of a p-variable (p>k) data set that are optimal under some
criterion. Search algorithms include a full search algorithm "leaps",
based on Furnival and Wilson's leaps and bounds algorithm, and three
random search algorithms: a genetic algorithm ("genetic"), a simulated
annealing algorithm ("anneal") and a modified local search algorithm
("improve"). Previous versions of subselect had three criteria for
optimality ("rm", "rv" and "gcd"), based on exploratory data analysis
considerations. This new version adds four new criteria ("tau2",
"zeta2", "xi2" and "ccr12"), associated with the multivariate linear
hypothesis and the well-known test statistics for that hypothesis
(Wilks' lambda, Roy's largest root, Pillai's trace and the
Hotelling-Lawley statistic). Subselect 0.9-99 also includes two
functions that prepare the arguments for the search functions
associated with the new criteria in the most frequent contexts of
application of the multivariate linear hypothesis - linear regression
(function "lmHmat") and linear discriminant analysis ("ldaHmat") - as
well as another function ("glhHmat") that prepares these arguments for
an analysis associated with a hypothesis specified by the user. The
package documentation includes numerous examples.  

Feedback is appreciated.

Best regards,
Jorge Cadima


-- 
Jorge Cadima
Departamento de Matemática
Instituto Superior de Agronomia
Universidade Técnica de Lisboa
Tapada da Ajuda
1349-017 Lisboa
Portugal

e-mail: jcadima at isa.utl.pt
phone: +351 21 365 3317
fax: +351 21 363 07 23