The results of the initial estimation of the first-order factor model did not reveal a
satisfactory results with a x 2 of 43.559 (df ¼ 24) significant at p , 0.009 level. Other fit
indices provided a moderate level of fit (RMSEA ¼ 0.069; GFI ¼ 0.948; CFI ¼ 0.936).
After analyzing the modification indices, we learned that we would achievea significantly better fit if we were to allow two of the error terms to be correlated. Based
on this information, we allowed the correlations between the errors of indicator variables
e2ˆ !e4 and e1ˆ !e7. Inspection of the Table IV shows the scales that were used
in the measurement model. The scales that needed to co-vary were 2.2 for marketing
strategy (e2) and 3.1 for information strategy (e4), respectively. These scales correspond
to “reduce complexity” and “coordination/control of channels”. Similarly, scales 2.1 for
marketing strategy (e1) and 1.1 for process strategy (e7) were correlated. These scales
correspond to “coordination of distribution” and “maximum efficiency”, respectively.
As shown in Figure 2, this resulted in significantly improved model fit (x 2 ¼ 31.058;
p ¼ 0.095;RMSEA ¼ 0.049; GFI ¼ 0.962; CFI ¼ 0.970). Non-significantx 2 as well other
fit indices indicated a good fit between the model and the data. Furthermore, the results