14 Florida Agricultural Experiment Stations
where
vk- = Logarithm of the quantity (in ounces per hundred cus-
tomers) of orange concentrate purchased in store i,
week k, age j 11
i= 1...10
k=1...9
j 1, 2, 3
p; = Logarithm of regression constant
= Logarithm of effect of age j
Ai = 1 when r = j, zero otherwise 12
6, = Logarithm of effect of store i
so = 1 when p = i, zero otherwise
'k-_i = Logarithm of effect of week k
w, = 1 when t = k, zero otherwise
aI = Logarithm of effect of interaction of store i and age j
Li P = 1 when p = i and r = j, zero otherwise
q = Regression coefficients
= Logarithm of price in age 1, zero otherwise 13
X;k.-j Logarithm of price in age 2, zero otherwise
X ,ki_= Logarithm of price in age 3, zero otherwise
"k-- Random component.
An observation is completely "located" if i and k are given, but j is
carried to identify the "age" of price.
Variables A, S, W and L were introduced to allow each observation
to be expressed in terms of all of the parameters of the model. Writing
the model in this way is helpful in developing the least-squares normal
equations for estimating purposes.
The weekly purchases of all brands and can sizes combined were
expressed as a function of the price of a 6-ounce can of Brand B concentrate.