For low complexity cases, full factorial designs (FF) are generally
considered as the most appropriate design. However, the FF
approach is rather conservative, and very labor-intensive and
costly, certainly when a high number of variables and/or an
extended range of levels is considered. Recently, Mertens et al.
(2012) showed that a Latin-square design (LS) is an advisable
alternative to full factorial design for complex cases, i.e., when
a high number of environmental factors and/or a high number of
levels is considered. Second order experimental designs like
BoxeBehnken and Central Composite Design are specifically useful
to identify curvatures and are, therefore, best applied for response
surface modeling and better not used for square-root-type models
(Mertens et al., 2012).
For low complexity cases, full factorial designs (FF) are generally
considered as the most appropriate design. However, the FF
approach is rather conservative, and very labor-intensive and
costly, certainly when a high number of variables and/or an
extended range of levels is considered. Recently, Mertens et al.
(2012) showed that a Latin-square design (LS) is an advisable
alternative to full factorial design for complex cases, i.e., when
a high number of environmental factors and/or a high number of
levels is considered. Second order experimental designs like
BoxeBehnken and Central Composite Design are specifically useful
to identify curvatures and are, therefore, best applied for response
surface modeling and better not used for square-root-type models
(Mertens et al., 2012).
การแปล กรุณารอสักครู่..
