[cadynce] Trying to Predict Worst-Case End-to-End Times from CPU Utilizations

[Alan F. Karr]

Colleagues,

This is an initial and not notably successful attempt to do the 
worst-case analogue of what we did previously for average end-to-end 
times. It is NOT meant to be either definitive or prescriptive: it is 
just an exploration of relationships between what we are measuring in 
the GT4 tests and the model-derived CPU utilizations. There is little 
doubt that the linear regression model is not appropriate: there is 
neither domain nor statistical science behind it.

All this being said, the results (attached and to be posted) are that 
for 8 of 10 appstrings the R^2 is (statistically significantly) 
non-zero. To us, at least, the negative value of the e1 coefficient for 
appstring G defies explanation or interpretation.

--- Alan

-- 
******************************************************************
* Alan F. Karr, Director                     * Tel: 919.685.9300 *
* National Institute of Statistical Sciences * FAX: 919.685.9310 *
* 19 T. W. Alexander Drive (FedEx/UPS)       * karr at niss.org     *
* P.O. Box 14006 (USPS)                      * www.niss.org      *
* Research Triangle Park, NC 27709-4006      *                   *
******************************************************************

-------------- next part --------------
A non-text attachment was scrubbed...
Name: regression-max.ppt
Type: application/vnd.ms-powerpoint
Size: 187904 bytes
Desc: not available
Url : http://list.isis.vanderbilt.edu/pipermail/cadynce/attachments/20070523/d94d3bb2/regression-max-0001.ppt