A.
Colin
Cameron and Pravin K. Trivedi
BOOK CORRECTIONS - Latest update January 7, 2008.
Changes below dated July 26, 2007 or earlier have
been made in the second reprinting of the book.
Misspellings and Punctuation errors are given
separately: See Misspellings
and
Punctuation
| Page |
Date Posted |
Correction or Addition |
| p.xiii |
11/6/2006 |
Table of Contents Section 23.3 should be page
792 (and not 762) |
| p.9 |
11/6/2006 |
Eleven lines down should be
"If D is found" (and not "d") |
| p.56 |
11/27/2009 |
In (3.7) left-hand side should
by y_i1 with superscript 0 and not superscript 1 [George Leckie,
Birstol] |
| p.69 |
11/6/2006 |
Last line should be "ln w" (and not "y") |
| p.73 |
1/4/2007 |
Line before proposition 4.1 delete "and
uncorrelated" [Neal Beck, NYU] |
| p.85 |
2/18/2006 |
Bottom line should be
"censored models (see Section 16.9.2)." [Jeff Smith, Michigan] |
| p.89 |
3/30/2006 |
Third last line should be "q = 0.1, 0.5, and
0.9" and not "q = 0.1, 0.2, ..., 0.9 and
the regression line" [James MacKinnon, Queen's] |
| p.101 |
11/6/2006 |
Equation (4.52) should end with (X'Z)^-1
(and not (Z'X)^-1). [Tom Wansbeek, Groningen] |
| p.101 |
11/6/2006 |
Bottom line should be "see Section 6.5.4"
(and not 6.5.6) [Raymond Guiteras, MIT] |
| p.103 |
7/26/2007 |
End first paragraph should be
"see Section 6.5.4" (and not 6.54,.5) [Fernando Martel, NYU] |
| p.113 |
5/27/2005 |
Exercise 4-2 part (b) should be Hence directly
obtain a consistent estimate of the variance of μ_hat (and not Hence directly obtain the variance of y_bar) |
| p.114 |
11/6/2006 |
Exercise 4-4 parts (c)-(e) have several typos.
See mmaex04_4.pdf. [Tomek Katzur,
Groningen] |
| p.114 |
6/9/2005 |
Exercise 4-7 parts (d)-(f) need to be replaced.
See mmaex04_7.pdf. |
| p.164 |
6/9/2005 |
Exercise 5-1 is correct but the function is
close to linear. A better example uses E[y|x]=exp(0+0.04x)/[1+exp(0+0.04x)]. |
| p.165 |
6/9/2005 |
Exercise 5-7 part (c) is ML estimation (delete
the word NLS). |
| p.168 |
3/3/2006 |
Second line after first displayed equation
should be E[h(x)(y-g(x,β))] = 0 (and not E[h(x)(y-x'β)]) [Doug Miller,
UC-Davis] |
| p.173 |
5/4/2007 |
Six lines down should be G_0 in propostion 6.1
(and
not 5.1) [Ronald Bachmann, RWI Essen] |
| p.178 |
3/3/2006 |
Last displayed equation. The first and third
matrices are wrong and should be similar
to G_hat in (6.21). For these matrices the two terms being summed
over i should be x_i*x_i' and 3*utilde_i^2*x_i*x_i'. [Doug Miller,
UC-Davis] |
| p.181 |
5/6/2009 |
OIR statistic in (6.26) should
be multiplied by N. [Elias Einio, Helsinki]. |
| p.185 |
11/6/2006 |
Equation (6.40) at the end should have N that is
not bold. |
| p.189 |
3/6/2006 |
Theil's interpretation.
Change "suppose that in the reduced form model" to "Suppose that
in a first-stage model" [Doug Miller, UC-Davis] |
| p.189 |
1/7/2008 |
First line after (6.49) should
be "PI is a rxK matric (and not Kxr). [Umed Khilvatsho,
Groningen] |
| p.190 |
3/6/2006 |
Basmann's interpretation. Change "OLS reduced
form prediction" to "OLS first-stage predictions". [Doug Miller,
UC-Davis] |
| p.193 |
3/6/2006 |
Top line change "because to regressors" to
"because the regressors" [Doug Miller, UC-Davis] |
| p.194 |
11/6/2006 |
Line before (6.58) add "defined by" at the end
of the line. |
| p.196 |
1/7/2008 |
First line of nonlinear 2SLS section should have N^-1 that is not bold. [Umed Khilvatsho, Groningen] |
| p.197 |
1/7/2008 |
Second line of last paragraph should have PI' (and not PI) in two places. [Umed Khilvatsho, Groningen] |
| p.197 |
11/6/2006 |
Last displayed equation is correct but is easier
to read if z_i in the last term is moved to the beginning of the sum. |
| p.198 |
11/6/2006 |
Nine lines from bottom after Hence .... the "not
equal to sign" should be "=". Eight lines from bottom "= 0" should be "not equal to zero". |
| p.199 |
5/18/2005 |
In Table 6.4 NL2SLS column is 0.969, 0.041, 0.84
(and not 0.960, 0.046, 0.85) |
| p.199 |
11/6/2006 |
Third line from bottom delete "that" at the
start of the line. |
| p.202 |
11/6/2006 |
Seven lines down should be Section 10.2.5 and
not Section 16.2.5. |
| p.206 |
11/6/2006 |
In middle reference should be to (6.76) and not
(6.26). |
| p.213 |
11/6/2006 |
Seven lines from bottom G^2 should be G(G-1) to
allow for G normalization retrictions on B. |
| p.214 |
4/10/2007 |
Thirteen lines from top subscript at end of line
should be lower case g. [Jens Krueger, Friedrich Schiller University,
Jena]r |
| p.214 |
3/28/2006 |
In the displayed equation
for the 3SLS estimator the matrix OMEGA_hat should be SIGMA-hat. Same change two lines down and four lines down. SIGMA_hat = definition given for OMEGA_hat. |
| p.220 |
5/27/2005 |
Exercise 6-1 part (a) should be (y - exp(x'β))^2
(and not (y - (x'β))^2). and part (d) should be E[x(y - exp(x'β))] = 0 (so add = 0) |
| p.237 |
11/6/2006 |
fourth line down should be "Section 7.3.5"
(and not "3.5") |
| p.248 |
11/6/2006 |
At top the definition of W should have a
transpose after first h_hat. |
| p.248 |
11/6/2006 |
Three lines above (7.46) should
be Section 7.6.4 (and not 7.7.4). |
| p.254 |
11/6/2006 |
Eleven lines from bottom delete
"in generating y," |
| p.255 |
5/18/2005 |
In sections 7.8.1-7.8.3 the sample size was N=40
(and not N=30) |
| p.255 |
5/18/2005 |
Five lines from bottom should be z = (0.817 - 1)
/ 0.376 = -0.487 |
| p.256 |
5/18/2005 |
In section 7.8.3 the percentiles should be -1.89
and 1.80 (and not -2.62 and 1.83) |
| p.273 |
10/4/2007 |
Six and seven lines from bottom
the upper and lower bounds for H should not be multiplied by
N [Liz Cascio, Dartmouth] |
| p.277 |
5/6/2008 |
OIR statistic in (8.39) should
be divided by N. [Elias Einio, Helsinki]. |
| p.296 |
1/21/2007 | three lines from bottom should
be within 0.545 and not within 0.21 |
| p.300 |
11/6/2006 |
Section 9.3.4 uses the same data as on page 206.
So second paragraph should be "... using
the log hourly wage data of Figure 9.1" (and not "using using
generated data, a random sample of size 100 drawn from the N[0,25^2]
distribution. For the particular sample drawn the sample mean is 2.81
and the sample standard deviation is 25.27.") |
| p.310 |
4/10/2007 |
Equation (9.20) should have + b(x_0) (and
not - b(x_0)) [Jens Krueger, Friedrich Schiller University, Jena] |
| p.311 |
5/18/2005 | firth last line should be ....
for pedagogical reasons, nonkernel local regression ... |
| p.335 |
1/21/2007 |
Exercise 9-4(a) delete "density"
so reads "Obtain the kernel regression estimate for ..." |
| p.341 |
11/6/2006 |
First line of section 10.3 should be "The
leading method for maximizing" (and not "for obtaining") |
| p.343 |
4/10/2007 |
Three lines after (10.7) should have minus sign
before E[ ] and then replace "which" by "where E[ ]" [Jens
Krueger, Friedrich
Schiller University, Jena] |
| p.352 |
4/10/2007 |
Bibliographic Notes 5.3 should be 10.3
[Jens Krueger, Friedrich Schiller University, Jena] |
| p.362 |
7/26/2007 |
In third paragraph in three
places change sigmahat to s_thetahat [Paul Glewwe, Minnesota] |
| p.364 |
7/26/2007 |
Equation (11.6) change first minus sign to + (since t*[1-alpha/2] is usually negative) [Paul Glewwe, Minnesota] |
| p.370 |
4/10/2007 |
Third line above (11.13) should by i = sqrt(-1)
(and not i = -sqrt(1)) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.373 |
11/6/2006 |
Fifteen lines from bottom should
be "Subsampling is useful" (and not "Subsampling bootstraps
are useful") |
| p.380 |
11/6/2006 |
Thirteen and fourteen lines from bottom only one
set of parantheses is needed (use (Nh^-1 ) and not ((Nh^-1)) ) |
| p.393 |
5/6/2008 |
Nine lines down 5^20 or 95 trillion should be 20^5 or 3 million [Jorge Gonzalez-Chapela, Alicante] |
| p.398 |
4/10/2007 |
Second last line Section 12.4 should have extra
parenthesis after theta in two places [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.410 |
4/10/2007 |
First line should be "quasi-random sequences"
(and not psuedo-random sequences) [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.414 |
5/18/2005 |
Figure 12.3 vertical axis label should be f(x)
and kg(x) and legend should be kg(x) (and not g(x)) |
| p.414 | 5/4/2011 |
Four lines down delete ", x = r," |
| p.453 |
4/10/2007 |
Three lines from bottom should be 10,000
(and not 100,000) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.457 |
4/10/2007 |
Table 13.4 and the line below this table should
be evidence againsty H2 (and not against H1) [Jens Krueger,
Friedrich Schiller
University, Jena] |
| p.469 |
1/7/2008 |
Second line should be "Differentiating (14.5)" (and not 14.4) [Umed Khilvatsho, Groningen] |
| p.474 |
7/26/2007 |
In fourth last paragraph replace
"specified" by "classified" in theree places |
| p.488 |
11/6/2006 |
Exercise 14-2 in the second line should have no
x in the definition of LAMDA(z) [Tom Wansbeek, Groningen] |
| p.493 |
2/18/2006 |
First two lines should be "in the probability of
fishing from a beach, and an increase of 0.119, 0.080, and 0.068, respectively, in the probability of fishing from a pier, a private boat, and a charter boat." [Jeff Smith, Michigan] |
| p.495 |
11/27/2009 |
In the denominator of the last displayed equation replace subscript l with subscript k, so d_ikl and I_ikl [Christoph Braun, RWI, Essen] |
| p.501 |
1/7/2008 |
(15.16) on right-hand side should be bold 0 [Umed Khilvatsho, Groningen] |
| p.501 |
3/22/2006 |
(15.17) and two lines before (15.17) should have
minus sign before the expected Hessian. [Frank Windmeijer, Bristol] |
| p.515 |
4/10/2007 |
Four and five lines from bottom should be C=2 or
C=3 (and not c=2 or c=3) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.524 |
11/27/2006 |
In derivation of d2lnL/dbdb' there should
be a minus sign in the last two equalities [Tomek Katzur,
Groningen] |
| p.526 |
4/10/2007 |
Third last equation delete the final close
parenthesis. [Jens Krueger, Friedrich Schiller University, Jena] |
| p.507 |
1/7/2008 |
Item 4. in the bullet list should have superscript s in the sum (and not superscript t) [Umed Khilvatsho, Groningen] |
| p.508 |
3/22/2006 |
Possible error in (15.31) needs to be checked |
| p.535 |
11/27/2009 |
Last term in displayed equation in mddole of page should be (1-d_i)ln[1 - Sum_i etc.] not (1-d_i)ln[Sum_i etc.] [Andres Ramirez Hassan, AEAFIT Univ, Columbia] |
| p.540 |
3/22/2006 |
Figure 16.2 and the last paragraph of page 540
use an alternative definition of the Mills ratio. To be consistent with our preferred definition (see discussion on p.541), we should instead use Mills ratio = E[z|z>-c] = phi(c)/PHI(c) as in equation (16.22). This yields a Corrected Figure 16.2 and a corrected last paragraph of p.540 [Michael Boozer, Yale, and Andrew Jones, York] |
| p.541 |
11/6/2006 |
Line after (16.23) should be "see Section
16.10.1" (and not "see Exercise 16-1") [Tom Wansbeek,
Groningen] |
| p.541 |
11/27/2009 |
In last line of (16.24) should
end with [w+lamda(w)]^2} and not [w+lamda(w)]}^2 [Andres Ramirez
Hassan, AEAFIT Univ, Columbia] |
| p.550 |
11/6/2006 |
In second line of (16.39) the last term should
be E[y_2*|x,y_1*>0] (and not E[y_2*|x,y_2*>00
[Stephanie Scholder, Bristol] |
| p.559 |
11/17/2006 |
In the bottom line (a displayed eqaution) the
very final term should be ln p_j (and not ln p_k)
[Elizabeth Klee, Fed Board of Governors] |
| p.560 |
11/17/2006 |
In the first displayed equation the very
final term should be ln phat_j (and not ln phat_k)
[Elizabeth
Klee, Fed Board of Governors] |
| p.570 |
11/6/2006 |
Exercise 16-1 part (b) should be "all
200 observations" (and not "all 2,000 observations") [Tom
Wansbeek, Groningen] |
| p.597 |
4/10/2007 |
In the first line add an extra close parenthesis
at
the end of the expression. [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.598 |
4/10/2007 |
In the table change heading Duration to
Time [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.599 |
4/10/2007 |
In the table change heading Duration to
Time [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.605 |
11/6/2006 |
Fitth line should be "... RRUI and DRUI and 34
other regressors whose coefficients are not reported in Tables 17.7 and
17.8" (and not just "... RRUI and DRUI") [Dimitri Szerman,
IPEA, Brazil]. |
| p.644 |
4/10/2007 |
First displayed equation after (19.6) should be
Δt_j in numerator (and not Δt) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.672 |
7/18/2007 |
In Table 20.4: LC should have mean 2.384 and sd 2.042, IDP mean 0.260 and sd 0.439, and FMDE 4.030 and 3.471 [Markus Jochmann, Universite catholique du Louvain] |
| p.699 |
11/6/2006 |
In (21.2) add an error u_it at the end of the
right-hand side [Tom Wansbeek, Groningen] |
| p.710 |
1/7/2008 |
In Table 21.2 beta should not be bold [Umed Khilvatsho, Groningen] |
| p.711 |
4/10/2007 |
Seven lines from bottom should be standard error
is 0.009 (and not 0.09) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.725 |
1/7/2008 |
In Table 21.7 beta should not be bold [Umed Khilvatsho, Groningen] |
| p.733 |
4/10/2007 |
Ten lines from bottom in the within group
variation definition delete term +y_bar [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.734 |
1/7/2008 |
Third line of second paragraph beta should be bold [Umed Khilvatsho, Groningen] |
| p.734 |
1/7/2008 |
(21.45) delta should be bold [Umed Khilvatsho, Groningen] |
| p.735 |
1/5/2007 |
Formula (21.48) should only be used if there are
no time-invariant regressors. If there are time-invariant regressors,
adjustment is needed, such as using the pooled OLS estimator rather
than the within estimator in (21.48). [Tom Wansbeek, Groningen] |
| p.736 |
11/6/2006 |
In the first displayed equation
remove the transpose operator in all three places. |
| p.736 |
11/6/2006 |
In (21.52) remove the transpose operator in all
three places. And replace bold W_bar by bold e times bold w_i bar transpose. |
| p.745 |
11/6/2006 |
In (22.5) remove the transpose on W_N
[Tom Wansbeek, Groningen] |
| p.749 |
7/26/2007 |
In the first line Z_i is T x Tr (and not Tr x T) [Christoph Braun, RWI, Essen] |
| p.760 |
11/6/2006 |
The first displayed equation is
valid in general but justification using Section 6.3.7 needs the additional assumption that E[X_i|Z_i] is a linear combination of Z_i. Reason: In equation (6.24) we have r( ) = u_i so E[du_i / dbeta | Z_i] = E[-X_i | Z_i] [Tom Wansbeek, Groningen] |
| p.763 |
11/13/2006 |
In equation (22.32) should be t=2,...,T
(and not t=1,...,T) [Tom Wansbeek, Groningen] |
| p.764 |
11/13/2006 |
Third line from bottom the intial value should
be "y_i1" (and not "y_i0) [Tom Wansbeek, Groningen] |
| p.765 |
11/13/2006 |
In equation (22.34) should be t=3,...,T
(and not t=2,...,T) [Tom Wansbeek, Groningen] |
| p.766 |
11/13/2006 |
In last paragraph of section 22.5.3 it is
assumed that the initial value y_i0 is observed (whereas elsewhere in
the section it was assumed that y_i1 was the initial value). |
| p.785 |
4/10/2007 |
In (23.19) f(y_it| ...) and gamma to the
conditioning and g(.) should be g(alpha_i|eta) [Jens Krueger,
Friedrich Schiller University, Jena] |
| p.793 |
4/7/2005 |
Figure 23.1 axes labels are reversed. Vertical
axis is log(patents) and horizontal axis is log(R&D) |
| p.795 |
1/9/2007 |
In (23.35) left-hand side should also condition
on beta. [Tom Wansbeek, Groningen] |
| p.796 |
1/9/2007 |
In (23.37) and (23.38) left-hand side should
condition on capital bold X_i (not x_i) Also on pages 795/796 the ordering of conditioning variables varies. [Tom Wansbeek, Groningen] |
| p.796 |
1/9/2007 |
Two lines above (23.37) should be "Section
23.4.6" and not "Section 23.4.5" [Tom Wansbeek, Groningen] |
| p.796 |
1/9/2007 |
Second last paragraph should begin "First, it is
not meaningful to condition on" rather than "not possible" [Tom
Wansbeek, Groningen] |
| p.797 |
1/9/2007 |
Line above (23.40) should be "Section 23.4.6"
and not "Section 23.4.5" [Tom Wansbeek, Groningen] |
| p.799 |
1/9/2007 |
In (23.43) drop the second equality and in the
next equality replace Pr[y] with Pr[y, sum_t y_t = c. [Tom Wansbeek,
Groningen] |
| p.799 |
4/10/2007 |
in last equality in the middle of the page
delete the lone "+" in the denominator [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.803 |
4/10/2007 |
In (23.49) delete the last minus sign in this
expression. [Jens Krueger, Friedrich Schiller University, Jena] |
| p.839 |
4/10/2006 |
Second equality for SIGMA_c^-1 should not have
the inverse at the end. |
| p.839 |
4/10/2006 |
Formula for [I + aee']^(1/2)
should finish with ee' and not Mee'. |
| p.843 |
4/10/2007 |
Five lines from bottom should be (23.46)
(and not (23.34)) [Jens Krueger, Friedrich Schiller
University, Jena] |
| p.870 |
7/26/2007 |
In (25.34) change last plus sign
to minus sign and in last term denominator chage 1-PHI() to PHI()
[Onur Bayar, Boston College] |
| p.870 |
7/26/2007 | In (25.35) last term should be (sigma_1e - sigma_0e) and not (sigma_0e - sigma_1e) [Onur Bayar, Boston College] |
| p.878 |
11/6/2006 |
Three lines from bottom should be "the fixed
effect phi_i" (and not "the fixed effect alpha")
[Oscar Jorda, UCD] |
| p.891 |
11/12/2007 |
In (25.77) the coefficient of D_i is just gamma (and not gamma x alpha) [Valerio Filoso, Naples] |
| p.895 |
5/26/2005 |
Table 25.6 footnote b drop RE74*RE75 from the
list of regressors |
| p.895-6 |
10/29/2007 |
Standard errors in Tables 25.6
and 25.7 based on bootstrap are invalid as the bootstrap fails for
matching estimates of ATET; see A. Abadie and .W. Imbens (2006), "On
the Failure of the Bootstrap for Matching Estimates." Instead use the
variance formulae in A. Abadie and G.W. Imbens (2006), "Large sample
properties of Matching Estimators for Average Treatment Effects,"
Econometrica, 235-268. These can be implemented using Stata add-on
nnmatch. |
| p.904 |
11/14/2006 |
Two lines after equation (26.9) should be "...
essentially the same as (26.8)" (and not (26.9)) [Tom
Wansbeek, Groningen] |
| p.933 |
11/6/2006 |
In step 4 the variance should be sigma^2 times
the identity matrix (and not just sigma^2) |
| p.937 |
11/6/2006 |
In Table 27.4 Header should be "with 25%
missing" (and not "with 10% missing") |
| p.946 |
12/31/2006 |
First line after equation (A.3) should be
"Amemiya (1985, p.89)" (and not p.79) [Yoshitaka Tsuruta,
Tsukuba]. |
| p.947 |
4/19/2006 |
Equation (A.6) add " = 1 " so
that " Pr[lim b_N = b] = 1 " [Richard Green, UCD]. |
| p.954 |
11/6/2006 |
Three lines from end of Section
A.6.4 should be root-N(X_bar_N - mu) converges in distribution to
N[0, sigma^2]. |
| p.966 |
4/10/2007 |
Add Brown, B.W., and W.K. Newey (2002),
"Generalized Method of Moments, Efficient Bootstrapping, and Improved
Inference," Journal of Business and Economic Statistics, 20,
507-517. [Jens Krueger, Friedrich Schiller University, Jena] |
| p.970 |
4/10/2007 |
Add to references Cragg, J.G. (1983), "More
Efficient Estimation in the Presence of Heteroscedasticity of Unknown
Form," Econometrica, 51, 751-63. [Jens Krueger, Friedrich
Schiller University, Jena] |
| p.970 |
4/10/2007 |
Add to references Davidson, R, and J.G.
MacKinnon (2000), "Bootstrap tests: How many bootstraps?" Econometric
Reviews, 19, 55-68.
[Jens Krueger, Friedrich Schiller University, Jena] |
| p.978 |
4/10/2007 |
Add to references Heckman and Vytlacil (2002).
[Jens Krueger, Friedrich Schiller University, Jena] |
| p.993 |
11/6/2006 |
Add to the references Smith, A.A.
(1993), "Estimating Nonlinear Time-Series Models Using Simulated
Vector Autoregressions", Journal of Applied Econometrics, 8, S63-S84. |
| p.999 |
3/15/2007 |
Add p.341 to Dorsey, R.E. entry |
| p.1002 |
3/15/2007 |
Add p.341 to Mayer, W.J. entry |