Microsoft Office Tutorials and References
In Depth Information
Exhibit 3.25 Summary output for product E quarter 4
variables. For the quarter 4 regression model in Exhibit 3.25, the value is larger
(0.00845293), yet there is likely to be a signiﬁcant association between X and Y.
The smaller the Signiﬁcance F the better the ﬁt.
There are many other important measures of regression ﬁt that we have not dis-
cussed for time series errors or residuals—e.g. independence or serial correlation,
homoscedasticity, and normality. These are equally important measures to those we
have discussed and deserve attention in a serious regression modeling effort, but are
beyond the scope of this chapter.
Thus far, we have used data analysis to explore and examine our data, taking
what we can from each form of analysis and adding whatever is contributed to our
overall insight. Simply because a model, such as regression, does not ﬁt our data
does not mean that our efforts have been wasted. It is still likely that we have gained
insight: this is not an appropriate model and there may be indicators of an alternative
to explore. It may sound odd, but often we may be as well informed by what doesn’t
work, as by what does.