28 Florida Agricultural Experiment Stations
As a precautionary measure, a three-way check was performed
to establish the accuracy of computations leading to estimates
of the model parameters. Estimated values and residuals were
computed for the 90 observations of concentrate purchases.27
The following algebraic identities were then used to verify com-
putations:
f (Yik-- jIki) = REMAINDER (8) = ik j- (TOT)SSR,
where
(TOT)SSR = 4o Yik-i- [;Y;kij i+ k ;I-Y.
u kk -'-i F l
+ I [ Yiki-] + r[ bp : Yk-i-pk-i-"
Independent computations of these three sums of squares yielded
SY(Yik-i k-j-)2 = 0.149
REMAINDER (8) =0.149
,2
f ZYik-- (TOT)SSR= 0.153
The close agreement among the three sums strongly sug-
gested that calculations performed in the analysis were free of
important computational errors.
ECONOMIC IMPLICATIONS OF THE ESTIMATING MODEL
Although interesting from the standpoint of methodology,
the study was primarily motivated by the need for information
essential to a clearer insight into economic problems of citrus
marketing. As a consequence, final attention was directed toward
translating the statistical results into economic concepts. How-
ever, because of the limited scope of the study, no claim is made
that the results have general application to "real world" problems.
the proportional reduction in variation due to regression. A more realistic as-
sessment of the fit of the model might be given by the following computation.
variation among original 88 observations = 1.668241
error sum of squares from model = 0.149083
reduction in sum of squares due to model = 1.519158
proportional reduction = 0.911
"- Estimates of these parameters were obtained by resorting to the nor-
mal equations of Appendix IV and the restrictions of Table 1, Appendix III.
"2 It should be noted that the residuals for the observations with missing
Y's were zero in each case.