1
$\begingroup$

What are some of the criteria for choosing one family over the other of experiment design methods ? How does this choice relate to the model of the response ?

I left factorial / fractional factorial out of the title because I see them as special cases of optimal experiment design when the number of runs/candidate points ratio is high. Am I correct ?

I'm specially interested in the following context:

More than 10, all fixed, multi level factors (i.e. categorical variables, with different number of categories), with constraints in some treatments.

Improve this question
$\endgroup$
3
  • $\begingroup$ Well, you will aso need to tell us what is thebgoal of this experiments $\endgroup$
    – kjetil b halvorsen ♦
    Commented Mar 20, 2018 at 16:58
  • $\begingroup$ These are numerical experiments (with non deterministic results), the goal is the comparison of different algorithms, and identification of the factors affecting performance. $\endgroup$
    – cladelpino
    Commented Mar 20, 2018 at 17:13
  • $\begingroup$ How costly are each run? Can you explain some of the factors? How large variance? ... Search this site for surrogate models or response surface models, for instance stats.stackexchange.com/questions/32322/… $\endgroup$
    – kjetil b halvorsen ♦
    Commented Mar 21, 2018 at 11:35

0

Reset to default

Browse other questions tagged or ask your own question.