The McGraw-Hill Series Economics essentials of economics brue, McConnell, and Flynn Essentials of Economics

Anderson–Darling
and the
Jarque–Bera tests
of normality. The reader is strongly urged to apply these or other tests
of normality to regression residuals. Keep in mind that in finite samples without the nor-
mality assumption the usual 
t
and 
F
statistics may not follow the 
t
and 
F
distributions.
We are left with Assumptions 1, 4, 5, 6, 7, 8, and 9. Assumptions 6, 7, and 8 are closely
related and are discussed in the chapter on multicollinearity (Chapter 10). Assumption 4 is
discussed in the chapter on heteroscedasticity (Chapter 11). Assumption 5 is discussed in
the chapter on autocorrelation (Chapter 12). Assumption 9 is discussed in the chapter
on model specification and diagnostic testing (Chapter 13). Because of its specialized
nature and mathematical demands, Assumption 1 is discussed as a special topic in Part 3
(Chapter 14).
For pedagogical reasons, in each of these chapters we follow a common format, namely,
(1) identify the nature of the problem, (2) examine its consequences, (3) suggest methods
of detecting it, and (4) consider remedial measures so that they may lead to estimators that
possess the desirable statistical properties discussed in Part 1.
5
Henri Theil, 
Introduction to Econometrics,
Prentice-Hall, Englewood Cliffs, NJ, 1978, p. 240. It must be
noted the assumptions of fixed 
X
’s and constant 
σ
2
are crucial for this result.
6
In passing, note that the effects of departure from normality and related topics are often discussed
under the topic of 
robust estimation
in the literature, a topic beyond the scope of this book. 
guj75772_ch10.qxd 12/08/2008 02:44 PM Page 318

---