Discussion:
Algebra, Eigen, Varimax. Precision of calculations.
(too old to reply)
p***@mail.com
2014-05-17 16:10:24 UTC
Permalink
Varimax rotation loadings matrix sizes of about 3000x3000, unless as a result of errors of machine completely loses ortogonality.
Please help with the rotation of large matrices.
Paul
Herman Rubin
2014-05-17 19:29:15 UTC
Permalink
Post by p***@mail.com
Varimax rotation loadings matrix sizes of about 3000x3000, unless as a result
of errors of machine completely loses ortogonality.
Post by p***@mail.com
Please help with the rotation of large matrices.
Paul
Multiplying by matrices of this size will lose SOME orthogonality,
as rounding errors add up, but if you are using "double precision"
as you should, it should be tolerable.

The real question is why are you using this method? The method
should be one which tackles the user's problem, and my experience
has been that the problem has not been correctly stated. It should
be stated using probability, but without any choice of statistical
method or methods.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
***@stat.purdue.edu Phone: (765)494-6054 FAX: (765)494-0558
paul analyst
2014-06-02 16:58:35 UTC
Permalink
I use this method and ask for precise calculations. Do you know have any programs or libraries that accurately count the variamx 3000x3000 matrix row?

Paul
Post by p***@mail.com
Post by p***@mail.com
Varimax rotation loadings matrix sizes of about 3000x3000, unless as a result
of errors of machine completely loses ortogonality.
Post by p***@mail.com
Please help with the rotation of large matrices.
Paul
Multiplying by matrices of this size will lose SOME orthogonality,
as rounding errors add up, but if you are using "double precision"
as you should, it should be tolerable.
The real question is why are you using this method? The method
should be one which tackles the user's problem, and my experience
has been that the problem has not been correctly stated. It should
be stated using probability, but without any choice of statistical
method or methods.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
Loading...