CCL: Randomization

[Yvonne Martin <yvonnecmartin+/-gmail.com>]

 Sent to CCL by: Yvonne Martin [yvonnecmartin===gmail.com]
 To do such a validation, separate the value to be predicted (Y) from
 the others, generate the same number of random numbers next to each of
 these Y values, sort these according to the value of the random
 number, and reassemble the dataset with the Y values to be predicted
 now dissociated from the others. Repeat this scrambling 100 or so
 times to see if your true model is superior to these random models.
 Yvonne Martin
 On Wednesday, November 10, 2010, Jon Heinz goeeoe*|*hotmail.com
 <owner-chemistry(a)ccl.net> wrote:
 >
 > Sent to CCL by: "Jon  Heinz" [goeeoe!A!hotmail.com]
 > Hello,
 >
 > I am trying to validate my QSAR models and I have read about Y-
 randomisation method. I do not understand what this technique means OR why it is
 important to carry it out. Are there any other suitable validation studies I can
 perform on my models, apart from cross validation?
 >
 > thank you
 > Jon Heinz
 > goeeoe[]hotmail.com
 > Portsmouth UK>       http://www.ccl.net/cgi-bin/ccl/send_ccl_message>;   
    http://www.ccl.net/cgi-bin/ccl/send_ccl_message>;   
    http://www.ccl.net/chemistry/sub_unsub.shtml>;   
    http://www.ccl.net/spammers.txt>;
 >
 >