Is it possible and valid to generate a response surface through augmentation of a fractionated factorial design, where all main effects and two factor interactions have no bias from any other main effects or two factor interactions (a resolution V design)?
$\begingroup$
$\endgroup$
10
-
3$\begingroup$ This is a little short for a question? Why do you think there would be a problem? $\endgroup$– kjetil b halvorsen ♦Mar 5, 2016 at 13:45
-
2$\begingroup$ Could you explain how this differs from your previous question? $\endgroup$– whuber ♦Mar 5, 2016 at 16:06
-
2$\begingroup$ It would help others looking at this site to say a bit about what a "resolution V design" is, to describe the type of underlying subject-matter question you are trying to solve, and the problems that you fear will arise in your proposed approach. $\endgroup$– EdMMar 5, 2016 at 16:30
-
$\begingroup$ There was a problem with the site and I had a response saying the question had been deleted (?). My issue is around what information could be lost, is there a higher likelihood of lack of fit, model fitting not always leading to good prediction. I am not a statistician, but I know some basics. $\endgroup$– BeerhunterMar 5, 2016 at 20:19
-
1$\begingroup$ Well, yes; that's quite standard: see e.g. itl.nist.gov/div898/handbook/pri/section3/pri3364.htm $\endgroup$– Scortchi - Reinstate Monica ♦Mar 7, 2016 at 21:51
|
Show 5 more comments