Heteroskedasticity Adjustment in Motel-Hotel Model
 The motel and hotel group had heteroskedasticity associated
with the number of rooms variable (NR). Thus, the variance
(,2) changed with the levels of NR, such that r2 = f(NR,). As a
result, the standard errors of the coefficients (on the independent
variables) will be biased. Confidence limits and tests of signifi-
cance using the calculated variances of the p's would then be in-
valid [14, p. 255]. It appeared r2, increased at an increasing rate
for increases in NR; hence it was posited 2, = r2 (NR )2 was a
reasonable proxy of the actual relation between the ,r2 and NR.
 Given the assumed relation o2, = r2(NRi)2, the appropriate
adjustment involves multiplying all variables by (1/NR,) [14,
p. 260]. The underlying models for Equation (3.3) and (3.4) in
the text (page 19) serve to illustrate the adjustment process. The
guiding model for Equation (3.3) was
 In (W,) = o + 3,ri + P2NNRk + p PR, + Ei. (A.9)
Equation (A.10) was derived by multiplying (A.9) by (1/NR;) or

 S(0 )+ )1 ( + 2 + P3 + Ei.
 NR X NRNR; 9\NR
 (A.10)

The coefficients estimated in (A.10) correspond directly to those
in (A.9) and can be used directly in the original model formula-
tion [14, pp. 260-261]. The standard errors of the adjusted co-
efficients did increase slightly. This was expected, as the bias on
the standard errors is negative when there is a positive relation
between o" and the variable of concern (in this case NR,) [14,
p. 255]. The t-tests for testing significance are valid after the
appropriate adjustment has been made [14, p. 255] and the as-
sumption regarding normality of the error term is retained. The
R' estimates for the adjusted model are also valid as the process
used to alleviate the problem did not remove the constant term
from the model.