4.5.4 Computing Q from Differences in Sums of Squares
As an alternative to the general formula for Q, equation 4.38, the sum of
squares for any hypothesis can be determined from the difference between
the residual sums of squares of two models. The current model, in the
context of which the null hypothesis is to be tested, is called the full model.
This model must include all parameters involved in the null hypothesis and
will usually include additional parameters. The second model is obtained
from the full model by assuming the null hypothesis is true and imposing
its constraints on the full model. The model obtained in this way is called
the reduced model because it will always have fewer parameters than the
full model. For example, the null hypothesisH_0: β_2 = c, where c is some
known constant, gives a reduced model in which β_2 has been replaced with
the constant c. Consequently, β_2 is no longer a parameter to be estimated.