The model suggested to explain relationships between the quantities is as follows:
The model reflects the beliefs that the relationships between the yields
and the temperature 
are approximately linear in
over the given range of temperature values. The intercept terms
indicate the yields for average temperature settings, whilst
the slopes of the regressions are given by 
. The models
incorporate error components 
. Separate runs of the
experiment are independent; however, in any particular run it is felt
that the error components will be correlated because slight aberrations
in reaction conditions or analytical procedures could simultaneously
affect both product yields. We will thus suppose that
are an uncorrelated sequence of error components with
expectation zero and variance 
; that 
are an uncorrelated sequence of error components with expectation zero
and variance 
; and that all pairs of error components
are uncorrelated except for
.
