Re: generate AR(1) data

[Mike Miller]

Heber--

Do you mean that you know the desired variance-covariance matrix for the
normal vectors?  If so, use the Cholesky factorization method.

If V is your symmetric toeplitz matrix and it is k-by-k, say, and you want
to make n normal vectors of length k, then do this:

randn(n,k)*chol(V)

That produces n multivariate normal vectors of length k having the
variance-covariance structure V.  You can also do this sort of thing:

randn(100,4)*chol(toeplitz([1 .5 .25 .125]'))

Check like so:

corrcoef(randn(10000,4)*chol(toeplitz([1 .5 .25 .125]')))

Mike



On Wed, 16 Jul 2003, Heber Farnsworth wrote:

> This is probably a simple question but I'm stumped and if someone knows
> the answer off the top of their head it would greatly speed up my code.
>
> I need to generate a large number of random (Gaussian) vectors that
> have an AR(1) correlation structure.  Obviously I could do this in a
> loop but this is slow.  I would rather rotate a vector of IID normal
> draws so that the result will have the right covariance matrix.  So
> what I need is the matrix square root of a symmetric toeplitz matrix
>
> 1   .5  .25  .125 ...
> .5   1  .5    .25 ...
> .25 .5  1   .5 ...
>
> etc (here the correlation is .5).  It seems inefficient to try to
> compute this square root numerically.  Isn't it known in closed form?
>
> Heber Farnsworth
>
>
>
> -------------------------------------------------------------
> Octave is freely available under the terms of the GNU GPL.
>
> Octave's home on the web:  http://www.octave.org
> How to fund new projects:  http://www.octave.org/funding.html
> Subscription information:  http://www.octave.org/archive.html
> -------------------------------------------------------------
>



-------------------------------------------------------------
Octave is freely available under the terms of the GNU GPL.

Octave's home on the web:  http://www.octave.org
How to fund new projects:  http://www.octave.org/funding.html
Subscription information:  http://www.octave.org/archive.html
-------------------------------------------------------------