- Original Message -
From: avraham.ad...@guycarp.com
Date: Thursday, June 18, 2009 10:54 am
Subject: RE: [R] Matrix inversion-different answers from LAPACK and LINPACK
To: Ravi Varadhan
Cc: r-help@r-project.org
> Thank you. One question, though. In the case where I have closed form
> f
Subject
RE: [R] Matrix inversion-different
-help-boun...@r-project.org] On
Behalf Of avraham.ad...@guycarp.com
Sent: Wednesday, June 17, 2009 6:11 PM
To: Douglas Bates
Cc: dmba...@gmail.com; r-help@r-project.org
Subject: Re: [R] Matrix inversion-different answers from LAPACK and LINPACK
I will be the first one to admit I may be doing s
-Original Message-
From: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On
Behalf Of avraham.ad...@guycarp.com
Sent: Wednesday, June 17, 2009 6:11 PM
To: Douglas Bates
Cc: dmba...@gmail.com; r-help@r-project.org
Subject: Re: [R] Matrix
am.ad...@guycarp.com,
r-help@r-project.org
06/17/2009 05:55 Subject
PM Re: [R] Matrix inversion-different
answers fro
On Wed, Jun 17, 2009 at 2:02 PM, Albyn Jones wrote:
> As you seem to be aware, the matrix is poorly conditioned:
>
>> kappa(PLLH,exact=TRUE)
> [1] 115868900869
>
> It might be worth your while to think about reparametrizing.
Also, if it is to be a variance-covariance matrix then it must be
positiv
As you seem to be aware, the matrix is poorly conditioned:
> kappa(PLLH,exact=TRUE)
[1] 115868900869
It might be worth your while to think about reparametrizing.
albyn
On Wed, Jun 17, 2009 at 11:37:48AM -0400, avraham.ad...@guycarp.com wrote:
>
> Hello.
>
> I am trying to invert a matrix, and
Subject
RE: [R] Matrix inversion-different
answers from LAPACK a
PM cc
Subject
RE: [R] Matrix inversion-different
cc
Subject
RE: [R] Matrix inversion-different
Avraham,
You can make LAPACK work by doing the following:
Hinv[, 1] <- solve(qr(PLLH, LAPACK=TRUE), c(1,0))
Hinv[, 2] <- solve(qr(PLLH, LAPACK=TRUE), c(0,1))
Here is an example:
H <- matrix(runif(4), 2, 2)
H <- H + t(H)
Hinv <- solve(qr(H)) # this is the correct inverse from LINPACK
Hinv1 <-
Hi Avraham,
I think this is a bug in solve() and qr.solve().
The structure of the QR object produced by LINPACK and LAPACK are different.
In fact, the help page for qr says:
"qr a matrix with the same dimensions as x. The upper triangle contains the
R of the decomposition and the lower triang
On Wed, Mar 5, 2008 at 7:43 AM, Duncan Murdoch <[EMAIL PROTECTED]> wrote:
> On 3/5/2008 8:21 AM, gerardus vanneste wrote:
> > Hello
> >
> > I've stumbled upon a problem for inversion of a matrix with large values,
> > and I haven't found a solution yet...
Someone with experience in numerical l
Sorry, I meant to send this to the whole list.
On Mar 5, 2008, at 8:46 AM, Charilaos Skiadas wrote:
> The problem doesn't necessarily have to do with the range of data.
> At first level, it has to do with the simple fact that dfdb has
> rank 6 at most, (7 at most in general, though in your ca
On 3/5/2008 8:21 AM, gerardus vanneste wrote:
> Hello
>
> I've stumbled upon a problem for inversion of a matrix with large values,
> and I haven't found a solution yet... I wondered if someone could give a
> hand. (It is about automatic optimisation of a calibration process, which
> involves the
Ben Domingue asks:
> I am trying to invert a matrix for the purposes of least squares. I
> have tried a number of things, and the variety of results has me
> confused.
Don't be.
> 1. When I try solve() I get the following:
> >Error in solve.default(t(X) %*% X) : system is computationally
> sing
Wang Chengbin wrote:
> I got the following error:
>
> a = read.csv("mat.csv")
> b = as.matrix(a)
> tb = t(b)
> bb = tb %*% b
> dim(bb)
> ibb = solve(bb)
> bb %*% ibb
>
>
>> ibb = solve(bb)
>>
> Error in solve.default(bb) :
> system is computationally singular: reciprocal condition number
Hello Wang
matrix bb is symmetric positive semidefinite, so
algebraically the eigenvalues are nonnegative.
I would use
bb <- crossprod(b)
to calculate bb (faster and possibly more accurate)
Look at eigen(bb,TRUE,TRUE)$values
(see ?eigen for the meaning of the arguments) to see how
many very s
18 matches
Mail list logo
