after all that there only remains to give some reference - for example Luc Devroye, Non-Uniform Random Variate Generation, available online at http://cgm.cs.mcgill.ca/~luc/rnbookindex.html (check Chap 11.2).

after all that there only remains to give some reference - for example Luc Devroye, Non-Uniform Random Variate Generation, available online at http://cgm.cs.mcgill.ca/~luc/rnbookindex.html (check Chap 11.2).

May be, L, you want to look up some probability books, but if I remember correctly then the following, more generally, holds: If your correlation matrix writes as C = A * A^t and W is a vector of n (n=100) independend normals (which as above they are), each of the desired lenght (you want about 100), then A * W is what you want. If your C = rho * Id is trivial (may be that is what you have in mind), A is obvious. Otherwise Maple might give you a Cholesky decomposition. For computing such a n-dim sample you might try Stephen's approach as shown by Alec.

May be, L, you want to look up some probability books, but if I remember correctly then the following, more generally, holds: If your correlation matrix writes as C = A * A^t and W is a vector of n (n=100) independend normals (which as above they are), each of the desired lenght (you want about 100), then A * W is what you want. If your C = rho * Id is trivial (may be that is what you have in mind), A is obvious. Otherwise Maple might give you a Cholesky decomposition. For computing such a n-dim sample you might try Stephen's approach as shown by Alec.

you could try to use a simple C++ version to pass from a global to a local situation (and then use Maple10 calling NAG to go on): http://iridia.ulb.ac.be/~fvandenb/download.php?id=19 gives a code which seems to connect easily by external calling (but i have not done that so can not judge the quality) another way would be to try simulated annealing (which should be possible in Maple) to find start values for local optima a little bit off the topic: i would like to see examples for Maples optimizer for numerical procedures (i e the objective function will be given through external libraries), i miss a clear guide for that. Especially on handling the various variants for constraints - would call the docu for users a kind of draft or 'beta' version

Aaahhh, thx ... in my context the message was irritating for me. On my wishlist: floating point support for special functions ...

Aaahhh, thx ... in my context the message was irritating for me. On my wishlist: floating point support for special functions ...

Alec, Nina, Thank you, I will try it with MSVC over the next days ...

This may be so (without any warnings) if you do not have a working web connection (firewall or modem or ...)

Nina, Just as a thought/wish/suggestion: If you are revisting that then why not provide the solution in a form which would allow to use the Compile facilty or the included Watcom compiler directly? May be as a separate section in that sheet (since it is more towards coding than financial application)?

Alec, Thx - but unfortunately not, it has just the same file name, but itself requires more from the lib there, for example its line 6 wants sir/layer0/random.h ... so the simpliest is if Maple completes the files :-)

through the menu support/downloads at www.maplesoft.com one finds ftp://ftp.maplesoft.com/pub/maple/downloads/ and there it is ... even on a fast connection it needs some time for downloading

On WIN ME under _java interface_ i get informed that a newer one is available. The popup wants to load it (or not). As I have a modem here and a light-wave cable connection at a place where my Maple is not installed I would prefer to have the URL only (to burn the thing onto a CD). Is this possible? Or is it only an incremental update of some moderate size?

As the params are missleading due to a typo i had this is what i meant www.axelvogt.de/maplekram/AJ.zip where i have not cared for multiple data. A compiled version would certainly speed up the calculation of chi^2.

As the params are missleading due to a typo i had this is what i meant www.axelvogt.de/maplekram/AJ.zip where i have not cared for multiple data. A compiled version would certainly speed up the calculation of chi^2.