Many experiments have been analyzed weekly, comparing
treatments and applying multiple range tests
(Harms et al., 1999; Leigh et al., 2010; Rama Rao et al.,
2011). It is not appropriate to apply such models to egg
production data from different levels of nutrients for 2
main reasons. First, considerable experimental power is
lost by analyzing segments with low degrees of freedom.
Second, regression inherently assumes that there are
unique values for responses to each level of input for
linear models: 2 for quadratic models, 3 for cubic, and
so on. It is not necessary to test to see if responses are
different at different level of input because the model
used assumes they are.