superdesolator / nortara Goto Github PK

Hey,
at the moment the NORTARA package is not available for R 4.X.X via CRAN.
If the NORTARA package is installed manually on R 4.X.X, it does not work. Running the example for NORTARA::genNORTARA leads to the Error:

Error in if (class(chol_decompose_val) == "try-error") { :
the condition has length > 1

Hi, I am using Nortara for producing high-dimensional vectors from a given correlation matrix.

I have noticed that, for small dimension D (e.g., D=10), generating N sample vectors requires a few minutes (we tested different N values, from 1000 to 10^6; the requested time was always the same).

But for large D values (e.g., D=28), Nortara becomes increasingly resource-demanding with N, terms of CPU time and occupied memory. We performed tests with N=200, 500, 1000.. this latter one never came to an end.

Is this behavior expected?

In a paper by Ghosh and Henderson dated 2003 (see ref. below), it is stated that the NORTA method becomes more and more ineffective as D increases.
The same authors propose a fix to NORTA, in order to solve this issue.

Did you implement Nortara taking into account such an issue (and fix)?

The aforementioned paper is:

https://people.orie.cornell.edu/shane/pubs/NORTAHighD.pdf