I started writing some backend regression routines, with the intent of
creating a regression procedure. Since least-squares fitting is a
cliche in statistics, these routines should be as general-purpose as
possible, so I created a library in lib/regression.  I haven't gotten
far yet: just sweep.c, regression.h and Makefile.am, but
it does compile to libregression.a.

Before I go any farther, is this an acceptable approach?
libregression.a will know almost nothing about PSPP, but different
PSPP procedures will be able to call its functions 'easily'.  There
are a lot of computational routines for linear models.  If possible,
such routines should be separated from the particularities of the many
PSPP procedures that eventually will need those computations. (A
semi-independent regression library might be useful in other
programs.)

If this sounds okay, I'll mail lib/regression to the list. If it's not
okay, what kind of design do you want?

-Jason

-- 
[EMAIL PROTECTED]
SDF Public Access UNIX System - http://sdf.lonestar.org


_______________________________________________
pspp-dev mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/pspp-dev

Reply via email to