Sebastien,

We do not use these codes (generalized_cholesky and generalized_cholesky2) since years. These routines were called in Dynare 3 when (minus) the hessian matrix computed after the optimization (for the posterior mode) was not positive definite. I think we can remove these routines...

Best, Stéphane.

Sébastien Villemot sebastien.villemot@ens.fr writes:

Dear Samik,

Samik Raychaudhuri samikr@gmail.com writes:

...

I was doing some research on generalized Cholesky decomposition (from the Schnabel and Eskow paper), and stumbled across your GPL'd code of Dynare at: https://www.dynare.org/trac/browser/matlab/ generalized_cholesky2.m While testing the code, it seemed that the code does not decompose properly for any symmetric positive definite matrix greater than 3X3. Are you aware of this problem, or am I missing something here?

Thanks for your feedback.

We are not aware of any problem in our code, but of course we cannot rule out a bug.

It would be very useful for us if you could provide us with an example to replicate the problem: which input matrix you used, which output you got and which output you were expecting.

Best,

Hi Sebastian, Thanks for your reply. I have a test case for a simple 4X4 matrix which shows the difference of output between the matlab chol() method and the generalized_cholesky2() method. The result matrix are so different from the two methods that I am inclined to believe that I am doing something wrong, or calling the method in an unexpected way. I did go through the Dynare code and have a few comments. If you indeed think that there is problem in Dynare code, I will be happy to pass on the comments. Please find the log attached. Thanks. -Samik

On 11/30/2011 3:17 AM, Sébastien Villemot wrote:

Dear Samik,

Samik Raychaudhurisamikr@gmail.com writes:

...

I was doing some research on generalized Cholesky decomposition (from the Schnabel and Eskow paper), and stumbled across your GPL'd code of Dynare at: https://www.dynare.org/trac/browser/matlab/ generalized_cholesky2.m While testing the code, it seemed that the code does not decompose properly for any symmetric positive definite matrix greater than 3X3. Are you aware of this problem, or am I missing something here?

Thanks for your feedback.

We are not aware of any problem in our code, but of course we cannot rule out a bug.

It would be very useful for us if you could provide us with an example to replicate the problem: which input matrix you used, which output you got and which output you were expecting.

Best,