[R] glm.fit to use LAPACK instead of LINPACK

[Ravi Varadhan]

Ted,

LAPACK is newer and is supposed to contain better algorithms than LINPACK.  It is not true that LAPACK cannot handle rank-deficient problems.  It can.

However, I do not know if using LAPACK in glm.fit instead of LINPACK would be faster and/or more memory efficient.

Ravi.

____________________________________________________________________

Ravi Varadhan, Ph.D.
Assistant Professor,
Division of Geriatric Medicine and Gerontology
School of Medicine
Johns Hopkins University

Ph. (410) 502-2619
email: rvaradhan at jhmi.edu


----- Original Message -----
From: Ted <tchiang at sickkids.ca>
Date: Thursday, October 22, 2009 10:53 am
Subject: Re: [R] glm.fit to use LAPACK instead of LINPACK
To: "r-help at R-project.org" <r-help at r-project.org>


> Hi,
> 
> I understand that the glm.fit calls LINPACK fortran routines instead of
> LAPACK because it can handle the 'rank deficiency problem'.  If my data
> matrix is not rank deficient, would a glm.fit function which runs on
> LAPACK be faster?  Would this be worthwhile to convert glm.fit to use
> LAPACK?  Has anyone done this already??  What is the best way to do this?
> 
> I'm looking at very large datasets (thousands of glm calls), and would
> like to know if it's worth the effort for performance issues.
> 
> Thanks,
> 
> Ted
> 
> -------------------------------------
> Ted Chiang
>   Bioinformatics Analyst
>   Centre for Computational Biology
>   Hospital for Sick Children, Toronto
>   416.813.7028
>   tchiang at sickkids.ca
> 
> ______________________________________________
> R-help at r-project.org mailing list
> 
> PLEASE do read the posting guide 
> and provide commented, minimal, self-contained, reproducible code.