Example 14.17 on page 683 has some problems.
Rao recommends using Box 14.8 to derive standard
errors and to approximate df for contrasts,
but I recommend deriving them each time.  The
answers from example 14.17 are wrong.  The
estimated contrast is right (thetahat=17.34)
but the standard error should be .361
with dfhat=3.1.

#2 on p.675 includes the error variance 
component in expressions where it doesn't
belong.

In the online data for problem 14.20, "wheat.dat",
the response for plant 1 in pot 3 from trt 4 is
given in Table 14.4 of the text as 14.4, but appears
as 14.1 in the datafile online.