eigen() error: R Version 1.6.1 on Mac OS X (PR#2550)
Dan: Do you by any chance have atlas installed? My guess is that when you compile it with fink it links the atlas lapack libs, but when you compile it yourself it links Apple's vecLib. I'll look into it. In the meantime, could you run otool -L on R.bin (the one you compiled yourself)? -Jeff
On Wed, 5 Mar 2003, Daniel E. Weeks wrote:
Actually, I only get the problem if I use R Version 1.6.2 as automatically installed by 'fink'. If I download the source directly myself for Version 1.6.2 from CRAN and compile it, then I get the proper eigenvalues. Seems strange to me - I thought that the source code in the fink repository should be identical to that at CRAN? -- Dan -- ------ Forwarded Message From: Daniel E. Weeks <dweeks@watson.hgen.pitt.edu> Date: Wed, 05 Mar 2003 16:09:48 -0500 To: Stefano Iacus <jago@mclink.it> Cc: <R-devel@stat.math.ethz.ch> Subject: Re: [Rd] eigen() error: R Version 1.6.1 on Mac OS X (PR#2550) Dear Stefano, Yes, I have updated to R Version 1.6.2 (2003-01-10), running it in a Terminal window under Darwin Kernel Version 6.4. And, while La.eigen() appears to work fine, I still have the same problem with eigen():
x
X1 X2 X3 X4 X5 1 3.240 2.592 2.592 2.592 2.592 2 2.592 3.240 2.592 2.592 2.592 3 2.592 2.592 3.240 2.592 2.592 4 2.592 2.592 2.592 3.240 2.592 5 2.592 2.592 2.592 2.592 3.240
La.eigen(x)$values
[1] 13.608 0.648 0.648 0.648 0.648
eigen(x)$values
[1] 13.123930 0.648000 0.648000 0.648000 -1.732930 I was hoping that updating to Version 1.6.2 would make my bug go away. But if you can not reproduce this bug, then what is it about my system that is different from yours? Thank you, -- Dan -- On 2/14/03 4:37 PM, Stefano Iacus at jago@mclink.it wrote:
Under Darwin R 1.6.2 and R-1.6.1 I get
sg
[,1] [,2] [,3] [,4] [,5] [1,] 3.240 2.592 2.592 2.592 2.592 [2,] 2.592 3.240 2.592 2.592 2.592 [3,] 2.592 2.592 3.240 2.592 2.592 [4,] 2.592 2.592 2.592 3.240 2.592 [5,] 2.592 2.592 2.592 2.592 3.240
eigen(sg)$values
[1] 13.608 0.648 0.648 0.648 0.648 as well as under Carbon R 1.6.1 and 1.6.2 I cannot reproduce your behaviour. stefano On Venerd?, feb 14, 2003, at 21:37 Europe/Rome, dweeks@watson.hgen.pitt.edu wrote:
Consider this matrix:
sg
X1 X2 X3 X4 X5 1 3.240 2.592 2.592 2.592 2.592 2 2.592 3.240 2.592 2.592 2.592 3 2.592 2.592 3.240 2.592 2.592 4 2.592 2.592 2.592 3.240 2.592 5 2.592 2.592 2.592 2.592 3.240 If I compute the eigenvalues of the 'sg' matrix using R Version 1.5.0 (2002-04-29) under Linux (or using Version 1.4.0 (2001-12-19) under Solaris), I obtain:
eigen(sg)
$values [1] 13.608 0.648 0.648 0.648 0.648 I also obtain the same eigenvalues under R Version 1.5.1 (2002-06-17) under Mac OS X 10.2.3. However, if I compute the eigenvalues of the 'sg' matrix using R Version 1.6.1 (2002-11-01) under Mac OS X 10.2.3, I obtain a different answer:
eigen(sg)
$values [1] 13.123930 0.648000 0.648000 0.648000 -1.732930
det(sg)
[1] 2.399354
det(sg,method="eigenvalues")
[1] -7.312317 Thanks, -- Dan --
______________________________________________ R-devel@stat.math.ethz.ch mailing list http://www.stat.math.ethz.ch/mailman/listinfo/r-devel
------ End of Forwarded Message
Jeffrey S. Whitaker Phone : (303)497-6313 NOAA/OAR/CDC R/CDC1 FAX : (303)497-6449 325 Broadway Web : http://www.cdc.noaa.gov/~jsw Boulder, CO, USA 80305-3328 Office: Skaggs Research Cntr 1D-124