Errata to 1st and 2nd printing of Nonparametric regression and
generalized linear models, by Green and Silverman

p.51: equation (3.41). Reinsch's motivation was actually the
reverse of what we say here - he wanted to minimise the roughness
penalty subject to an upper bound on the error sum of squares.

We are grateful to Carl de Boor (Madison) for pointing this out.  

p.34: The second and third displayed equations are incorrect:
      They should read (using TeX style formatting):

\bar{b}{i,i+1} = - L_{i+1,i}\bar{b}{i+1,i+1} - L_{i+2,i}\bar{b}{i+1,i+2} 
\bar{b}{i,i+2} = - L_{i+1,i}\bar{b}{i+1,i+2} - L_{i+2,i}\bar{b}{i+2,i+2}

We are grateful to Erik Leertouwer (Groningen) and Berwin Turlach (ANU)
for pointing this out.

p.46: In line 15, X_{ij} should be defined as $\beta_j(t_i)$.

p.97: The second to last displayed equation should have derivatives
  with respect to \beta not \eta on the left hand side, thus:
$$
\left\{ E \left( - \frac{\partial ^2 \ell}
{\partial \beta \partial \beta ^T} \right)
\right\} ^{-1} = \phi (X^TWX) ^{-1} .
$$

We are grateful to Art Owen (Stanford) for pointing this out.

------------------------------

Errata to 1st printing of Nonparametric regression and
generalized linear models, by Green and Silverman

p.68: In the theorem, it should be assumed that A and C are symmetric.

p.100, line -9: replace sentence by
       The off-diagonal elements are just -\alpha K_{ij}.

p.128, one third way down: display should read
           N g = S(z-X beta)
   Also, a non-iterative solution to (5.24) is available in this case.

p.131: The W's and u's -> The u's and W's

We apologise to M. F. Hutchinson for consistently mis-spelling his name.

There are a number of errors in grammar, and trivial typos.

All of these are corrected in the 2nd printing (May/June 1995).

------------------------------

Please keep advising us of any further errors you encounter.

P.J.Green@bristol.ac.uk
B.W.Silverman@bristol.ac.uk