Fixed Effects

Table 1. Type III Tests of Fixed Effects for unstructured covariance model
Source	Numerator df	Denominator df	F	Sig.
Intercept	1	347.004	5753.625	.000
COUPVAL	3	428.318	101.769	.000
SEQ	3	348.699	1.069	.362
CARRY	3	414.568	17.665	.000
WEEK	2	348.523	.269	.765

The difference in the tests of fixed effects is that the carryover effect is now also a statistically significant factor.

Figure 1. Estimates of fixed effects for unstructured covariance model

The values of the levels of the carryover effect are better differentiated, as well as statistically significant. With respect to the carryover effect of the 25 percent coupon [CARRY=4], the carryover effect of no coupon [CARRY=1] is about 5.86. That is, if sending no coupon this week has no effect on customer spending in the following week, then sending a 25 percent coupon will correspond to a 5.86 decrease in spending the following week.

Further, the standard errors of the estimates of the immediate coupon effects are drastically reduced, so that relative to the 25 percent [COUPVAL=4] level, all of the others are statistically significant.