Econometric Analysis, 6th Edition
ERRATA and DISCUSSION
Last updated March 14, 2010
Page Number
Date Posted
Errata
PAGE NUMBER
DATE
44
11/30/08
55
06/28/09
57
09/16/08
69
10/12/09
69,70,71
09/16/08
The analysis is based on 51 observations, not 52.
Coefficient -0.178397 should be -0.178395
Standard error 0.04377 should be 0.0147327
Standard error 0.07771 should be 0.0550265
Standard error 0.15707 should be 0.0551728
Standard error 0.10330 should be 0.0357706
t Ratio -3.377 should be -3.233
At the top of page 70,
Reported long run price elasticity -0.411361 should be -0.411358.
Reported long run income elasticity 0.970527 should be 0.970522.
Reported asymptotic variance 0.026368 should be 0.0263692.
The critical t value for 51-6 = 45 degrees of freedom 2.013 should be 2.014.
The 95% confidence interval for the long run price elasticity should be
-0.411358 +/- 2.014(0.152296) = [-0.718098,-.104618].
In matrix C in Example 4.8:
0.0147326 should be 0.0147327,
0.00109461 should be 0.00110072,
-0.0396227 should be -0.0396215,
0.0216259 should be 0.0216279.
The remaining results on page 71 in Table 4.8 are dependent on the simulation.
The first two columns of the table are the results reported
earlier. The coefficient estimates are correct. The two derived estimates,
f_{2} and f_{3}
should be as given above and in the table below. The standard errors are as
shown earlier and reported again in the table below. The simulation results
will be specific to the application. A program that can produce the particular
results shown below and can be reproduced (not the results in the text) using NLOGIT
or LIMDEP can be found at delta-method-and-simulation.txt
========================================================
Regression Estimate Simulated Values
Parameter Estimate Std.Error Mean Std.Dev.
========================================================
BETA2 | -0.069532 0.01473267 -0.070018 0.0147507
BETA3 | -0.164047 0.05502650 0.162375 0.0541288
GAMMA | 0.830971 0.04576345 0.833081 0.0450839
PHI2 | -0.411358 0.152296 -0.462782 0.205537
PHI3 | 0.970522 0.162386 0.966962 0.188948
? Example 4.7 in Detail.
? Create logged variables for the regression
sample;1-52$
create;logg=log(gasexp/(gasp*pop)) ; logg1=logg[-1]$
create;logpg=log(gasp);logi=log(income)$
create;logpnc=log(pnc);logpuc=log(puc)$
? Drop first observation
sample;2-52$
regress;lhs=logg;rhs=one,logpg,logi,logpnc,logpuc,logg1; printvc$
? Pick off specific coefficients
calc ; br2=b(2) ; br3=b(3) ; cr=b(6)$
calc; list ; phir2=br2/(1-cr) ; phir3=br3/(1-cr)$
calc; list ; g22=1/(1-cr) ; g26=g22*phir2 ; g33=g22 ; g36=g33*phir3$
matrix ; g=[0,g22,0,0,0,g26 / 0,0,g33,0,0,g36]$
matrix ; list ; ve=g*varb*g'$
calc;list;sef2=sqr(ve(1,1)) ; sef3 = sqr(ve(2,2))$
? The preceding is automated with
Wald ; labels=b1,b2,b3,b4,b5,gama ; start=b ; var=varb
; fn1=b2/(1-gama) ; fn2 = b3/(1-gama) $
calc;list;ct=ttb(.975,45)$
? Report the confidence interval.
calc;list ; lowerf2=phir2-ct*sef2 ; upperf2=phir2+ct*sef2 $
calc;list ; lowerf3=phir3-ct*sef3 ; upperf3=phir3+ct*sef3 $
?
? Krinsky and Robb Procedure Standard Errors
?
matrix;b2b3c=[br2/br3/cr]$
matrix;vb2b3c=[varb(2,2)/varb(2,3),varb(3,3)/
varb(2,6),varb(3,6),varb(6,6)]$
? Cholesky decomposition of asymptotic covariance matrix.
Matrix;c=chol(vb2b3c)$
? Results can be replicated.
Calc;ran(1234567)$
matrix;v=rndm(1000,3) ; vc = v*c'$
sample;1-1000$
crea;beta2=0;beta3=0;gamma=0$
namelist;bsim=beta2,beta3,gamma$
create;bsim=vc$
create;beta2=beta2+b(2)
;beta3=beta3+b(3)
;gamma=gamma+b(6)$
create;phi2=beta2/(1-gamma) ; phi3 = beta3/(1-gamma)$
dstat;rhs=bsim,phi2,phi3$
calc ; list; lowerf2=xbr(phi2)-ct*sdv(phi2)
; upperf2=xbr(phi2)+ct*sdv(phi2)$
calc ; list; lowerf3=xbr(phi3)-ct*sdv(phi3)
; upperf3=xbr(phi3)+ct*sdv(phi3)$
?
? Use empirical percentiles instead.
?
sort;lhs=phi2$
sort;lhs=phi3$
calc;list;phi2(26);phi2(975)$
Result = -0.983866
Result = -0.209776
calc;list;phi3(26);phi3(975)$
Result = 0.539668
Result = 1.321067
========================================================
Regression Estimate Simulated Values
Parameter Estimate Std.Error Mean Std.Dev.
========================================================
BETA2 | -0.069532 0.01473267 -0.070018 0.0147507
BETA3 | -0.164047 0.05502650 0.162375 0.0541288
GAMMA | 0.830971 0.04576345 0.833081 0.0450839
PHI2 | -0.411358 0.152296 -0.462782 0.205537
PHI3 | 0.970522 0.162386 0.966962 0.188948
86
09/16/08
98
09/16/08
101
09/16/08
matrix ; bvec=[-17.6992/0.000500633/0.000339424/2.86348/-0.00460571]$
As given on page 100, the data vector is
matrix ; xvec=[1/4.48/5.26/9.1396/204]$
The estimate s^{2} is 0.000873331. The predicted value is, therefore,
calc;list ; xb=xvec'bvec ; if=exp(xb + ssqrd/2)$ = 1875.746
not 1885.2 as reported in the example. The overestimation is closer to 9.0% rather than
9.54% as reported.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
113
06/28/09
116
09/16/08
119
09/16/08
124
05/19/09
125
09/16/08
125
07/25/08
126
09/05/08
In the same example, in the last paragraph on page 126, the chi squared reported
there, 158.753, was computed by comparing the "Preshock" coefficient vector to the
"1953-2004" coefficient vector, which is the wrong test. If the statistic is computed
to compare the "Preshock" to the "Postshock," the value is 502.34 with the same conclusion.
(Professor Noel Roy, Department of Economics, Memorial University of Newfoundland)
142
09/06/08
164
09/16/08
164
09/16/08
165
09/26/08
167
09/16/08
c^{2}[2]=n[(1/6)(m_{3}/s^{3})^{2}
+(1/24)((m_{4}/s^{4})-3)^{2}]. The value given of 482.12
should be 497.35.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
171
09/16/08
173
09/16/08
There is a notational inconsistency in the text. In (8-28), Z includes
the constant term. In the discussion on page 173, that overall constant term has fallen
out of Z. To resolve the inconsistency, a way to proceed is to change Z
to D in the 4th and 6th lines of the paragraph before (8-37), and then Z=
[i,D]. In the rest of the paragraph, all occurrences of G-1 are
changed to G (3 times), and as noted above, the summation in (8-37) runs from 1 to
G not from 2 to G.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
173,174
09/28/08
174
03/11/08
crea;d1=i=1;d2=i=2;d3=i=3;d4=i=4;d5=i=5;d6=i=6
;d7=i=7;d8=i=8;d9=i=9;d10=i=10;d11=i=11;d12=i=12
;d13=i=13;d14=i=14;d15=i=15;d16=i=16;d17=i=17
;d18=i=18$
name;D=d2,d3,d4,d5,d6,d7,d8,d9,d10,d11,d12,
d13,d14,d15,d16,d17,d18$
name;Z=d1,d$
regr;lhs=lgaspcar
;rhs=one,lincomep,lrpmg,D;res=e$
calc;ee=e'e$
crea;g=e*e/(sumsqdev/n)-1$
matrix;list;lm=.5*g'Z*ginv(Z'Z)*Z'g$
Matrix LM has 1 rows and 1 columns.
1
+--------------
1| 277.44275
(Fernando Martel, Department of Politics, New York University and
Professor Pedro Bacao, University of Coimbra, Portugal.)
193
09/16/08
202
10/14/09
209, 210
09/16/08
name;x=exp,expsq,wks,occ,ind,south,smsa,ms,union$
samp;all$
regr;lhs=lwage;rhs=one,x;pds=7;panel$
regr;lhs=lwage;rhs=one,x$
regr;lhs=lwage;rhs=one,x;pds=7;panel;fixed$$
matr;bf=b;vf=varb$
regr;lhs=lwage;rhs=one,x;pds=7;panel;means$
matr;bm=b(1:9);vm=varb(1:9,1:9)$
matr;d=bf-bm;vd=vf+vm$
matr;list;h=d'*ginv(vd)*d$
crea;expb=groupmean(exp,pds=7)$
crea;expsqb=groupmean(expsq,pds=7)$
crea;wksb=groupmean(wks,pds=7)$
crea;occb=groupmean(occ,pds=7)$
crea;indb=groupmean(ind,pds=7)$
crea;southb=groupmean(south,pds=7)$
crea;smsab=groupmean(smsa,pds=7)$
crea;msb=groupmean(ms,pds=7)$
crea;unionb=groupmean(union,pds=7)$
name;means=expb,expsqb,wksb,occb,indb,southb,smsab,msb,unionb$
regr;lhs=lwage;rhs=means,x,one;panel;pds=7;random$
matr;c=b(1:9);vc=varb(1:9,1:9)$
matr;list;hp=c'*ginv(vc)*c$
(Professor Pedro Bacao, University of Coimbra, Portugal and William Greene,
New York University.)
217
09/16/08
The robust standard errors reported in parentheses in the second column of values
were computed by making a degrees of freedom correction to (9-4); the resulting matrix
has been multiplied by (nT)/(nt-K) = 816/810.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
216,224,254
11/23/07
p_cap = hwy + water + util
This same correction must be made at the beginning of Example 9.12 on page 224
and at the beginning of Example 10.1 on page 254. The numerical results given
in the examples are based on the correct specification, that is, using private
capital, not aggregate public capital.
In the first line of Example 9.9, "Munell" should be "Munnell." In the listing
of state abbreviations for Tornado Alley, "AK"should be "AR" and "MS" should be
"MO." These latter two state abbreviations should be also be changed likewise in
the definition of "Central" (which is the same as Tornado Alley) on page 254.
(Professor Randall Campbell, Mississippi State University.)
226
4/03/08
259
12/06/07
259-265
11/23/07
The Breusch-Pagan LM statistic based on the correlation matrix shown on page 259 in
the text is 119.242, not 215.88 as reported in the text. However, this is not the
correct correlation matrix for the OLS residuals based on the data used in
the example. The correct correlation matrix is
GF MW MA MT NE SO SW CN WC
GF 1.0000
MW .1734 1.0000
MA .3989 .0633 1.0000
MT .4467 .6969 -.0158 1.0000
NE -.5473 -.2896 .1915 -.5372 1.0000
SO .5253 .4893 .2328 .3433 -.2411 1.0000
SW .4247 .1320 .6514 .1301 -.3220 .2594 1.0000
CN .7509 .3370 .3904 .4957 -.2979 .8050 .3465 1.0000
WC .1672 .5654 .2116 .5736 -.0576 .2693 -.0374 .3818 1.0000
The chi-squared statistic based on this matrix is 102.336. This is also the
value that should be given in Example 10.4 on page 265, where the 215.879
is repeated. The 48 state value of 2455.71 also given in Example 10.4 is incorrect (it is based on
p_cap instead of pc. The correct value is 2915.564
In Table 10.1, the OLS results for GF and for the Pooled data are incorrect. The correct
results are as follows:
GF 11.5669 0.002124 -2.0285 0.1005 1.3577 0.8049 -0.006562 0.01097 0.9970
(4.8502) (0.3012) (1.14883) (0.1243) (-0.8655) (0.1592) (0.003643)
Pooled 2.5159 0.2685 -0.1549 0.3123 0.2438 0.3122 -0.01900 0.04171 0.9959
(0.1159) (0.01759) (0.02805) (0.03334) (0.03189) (0.03052) (0.002285)
The remaining OLS results require the following corrections:
For MW, 1.7834 should be 1.7835
For MT, 0.2948 should be 0.2949 and -0.008143 should be -0.008144
For NE, 6.3783 should be 6.3784
For SO, -0.02040 should be -0.02041 and 0.9851 should be 0.9852
For CN, 0.9936 should be 0.9937
For WC, 9.1108 should be -9.1108
Because of the incorrect covariance (correlation) matrix, the FGLS results must all be replaced.
The correct values (based on the OLS residuals, using no iteration, and no degrees of freedom
correction in the covariance matrix of the residuals) are
GF 11.8735 -0.16721 -1.8161 0.1706 1.1440 0.9349 -0.003974 0.008413 0.9970
(2.02941) (0.1270) (0.4618) (0.05565) (0.3296) (0.08108) (0.001951)
MW 4.1692 0.06698 -0.1397 -0.1332 0.5323 0.5372 -0.01565 0.007625 0.9984
(1.00323) (0.08627) (0.1285) (0.06115) (0.1121) (0.08544) (0.002502)
MA -8.4614 0.3278 1.6885 0.5173 -0.3140 -0.3093 -0.02906 0.007975 0.9950
(2.1246) (0.1056 (0.3301) (0.06105) (0.1270) (0.1839) (0.004445)
MT 2.5230 0.1767 0.1227 -0.3669 0.03931 1.09375 -0.005851 0.01295 0.9940
(3.4487) (0.09170) (0.1142) (0.1733) (0.4629) (0.1704) (0.004203)
NE 3.1392 -0.09790 0.2099 0.2446 -0.3088 1.04987 -0.001048 0.006597 0.9986
(1.3723) (0.04677) (0.1704) (0.04875) (0.1202) (0.09656) (0.002644)
SO -13.1441 0.09109 0.9961 -0.5823 -0.2988 2.5935 0.02130 0.01719 0.9852
(7.6199) (0.1293) (0.7830) (0.13609) (0.2405) (0.4672) (0.008097)
SW -19.1347 -0.3251 3.1643 -0.09668 -1.7129 2.1875 0.01712 0.009914 0.9864
(2.8916) (0.09034) (0.9215) (0.2055) (0.5817) (0.4932) (0.008129)
CN 2.7917 -0.03071 -0.5043 -0.02158 0.1591 1.5380 0.006153 0.01345 0.9937
(0.8510) (0.09576) (0.1866) (0.1436) (0.1674) (0.1789) (0.004281)
WC -10.2819 0.04750 1.8131 0.6651 -0.4305 0.004467 -0.02914 0.008320 0.9895
(2.4115) (0.1088) (0.4492) (0.09973) (0.1909) (0.1798) (0.003666)
Pooled 2.2802 0.2365 -0.03934 0.1226 0.3124 0.3653 -0.01494 NA 0.9995
(0.02538) (0.002373) (0.004915) (0.004646) (0.005230) (0.007060) (0.0003400)
FGLS 2.2312 0.2991 -0.1320 0.2687 0.2467 0.3177 -0.01967 NA 0.9967
Het. (0.03643) (0.004196) (0.006546) (0.007369) (0.007622) (0.008832) (0.0005227)
In example 10.3, using the FGLS results above, the correct chi squared statistic is 6,429.85, rather than 10,554.77
given in the example. The chi squared value for the constant returns to scale test should be changed from 148.418
to 136.390. The 9 item discrepancy vector shown on page 203 should be
(-0.73381, -0.13633, 0.91025, 0.06581, 0.09767, 1.79957, 2.21725, 0.14053, 1.09964)
(Professors Randall Campbell, Mississippi State University and William Greene, New York University.)
280
4/30/08
287
10/14/09
321
03/28/09
325
09/22/08
388
09/28/08
394
09/28/08
400
3/03/08
402
10/14/09
436
09/22/08
In the listing of the second derivatives, the first 1/n before the brackets
should be deleted. The result is [-H]^{-1}=... In the actual values,
0.51203 should be -.51243; 0.01637 should be 0.01638.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
450
3/03/08
460
09/22/08
477
05/16/08
478,479
09/22/08
479
09/22/08
495
3/03/08
497
3/03/08
505,506
10/06/07
-4.7185 (2.3445) 15.6027 (6.974)
3.1517 (0.7943) 1.0000 (0.000)
-82.91605 -88.43626
0.0000 0.0000
0.0000 7.9145
-0.85570 -0.026941
-7.4592 -32.8987
-2.2420 0.66891
The values in the three matrices should be changed to
V = [5.499/-1.653,0.6309],
V_{E} = [4.900/-1.473,0.5768],
V_{B} = [13.37/-4.322,1.537]
In the confidence interval test, 3.1517 should be 1.1509 and [1.5942,4.7085] should be [1.5941,4.7076].
In the likelihood ratio test, 11.0465 should be 11.0404.
In the Wald test, 3.1517 should be 1.1509 and 2.70895 should be 2.73335
In the Lagrange multiplier test, 7.9162 should be 7.9145 twice, the matrix should be
[0.00995/0.26776,11.199] and the result of 15.687 should be 15.686. On the next page,
the value 5.1182 should be 1.2653.
Further, using the 17 internal digits of the computation with no external rounding, the following
shows the results
? Obtain Maximum Likelihood Estimates.
create;logy=log(y)$
maximize;labels=beta,ro ; start = -2,2
;fcn=-ro*log(beta+e)-lgm(ro)+(ro-1)*logy-y/(beta+e)$
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|BETA | -4.71850 3.65680241 -1.290 .1969 |
|RO | 3.15090** 1.23984807 2.541 .0110 |
+--------+-------------------------------------------------+
| Note: ***, **, * = Significance at 1%, 5%, 10% level. |
+----------------------------------------------------------+
? Verify that first derivatives sum to zero
calc;bm=b(1);rm=b(2)$
create;bi=1/(bm+e)$
create;gb=-rm*bi+y*bi*bi;gr=log(bi)-psi(rm)+logy$
calc ;list;sum(gb);sum(gr)$
Result = .000000
Result = .000000
? Compute second derivatives and expected second derivatives.
create ;gbb=rm*bi*bi-2*y*bi^3;grr=-psp(rm);grb=-bi$
create;egbb=-rm*bi^2$
namelist;g=gb,gr$
? Sum second derivatives to obtain Hessian and expected Hessian
calc;hbb=-sum(gbb);hrr=-sum(grr);hbr=-sum(grb);ehbb=-sum(egbb)$
+------------------------------------+
| Listed Calculator Results |
+------------------------------------+
HBB = -.855704
HRR = -7.459184
HBR = -2.241969
EHBB = -.877926
matrix;list;opg=
When the second coefficient is fixed at 1.00000 the following results are obtained.
maximize;labels=beta,ro ; start = -2,1 ; fix = ro
;fcn=-ro*log(beta+e)-lgm(ro)+(ro-1)*logy-y/(beta+e)$
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|BETA | 15.6027 10.0255472 1.556 .1196 |
|RO | 1.00000 ......(Fixed Parameter)....... |
+--------+-------------------------------------------------+
? Verify that first derivative sums to zero
calc;bm=b(1);rm=b(2)$
create;bi=1/(bm+e)$
create;gb=-rm*bi+y*bi*bi;gr=log(bi)-psi(rm)+logy$
calc ;list;sum(gb);sum(gr)$
Result = .000000
Result = 7.914484
? Compute second derivatives and expected second derivatives.
create ; gbb=rm*bi*bi-2*y*bi^3;grr=-psp(rm);grb=-bi ; egbb=-rm*bi^2 $
namelist;g=gb,gr$
? Sum second derivatives to obtain Hessian and expected Hessian
calc;list ; hbb=-sum(gbb);ehbb=-sum(egbb) ; opg=gb'gb$
HBB = .021662
EHBB = .022597
OPG = .009949
? Estimates of standard error
calc ;list;sqr(1/hbb) ; sqr(1/ehbb) ; sqr(1/opg)$
Result = 6.794363 (Actual second derivative)
Result = 6.652414 (Expected second derivative)
Result = 10.025547 (Sum of squares of first derivatives.)
(Professor William Greene, New York University and Xu Zhongmin, China.)
520
3/03/08
525
9/22/08
l_{WALD} =
a^'{[0,I](2(Z'Z)^{-1})[0,I]'}^{-1}a^
(Brian Gould, University of Wisconsin, Professor Pedro Bacao, University of Coimbra, Portugal.)
526
09/22/08
527
03/28/09
528
09/22/08
529
09/22/08
548
09/22/08
549,550
09/22/08
576
3/03/08
578
1/24/10
583
12/23/09
587
03/06/10
628
09/28/08
636
09/22/08
646
03/28/09
649
03/28/09
650,651
10/06/07
+--------+--------------+----------------+
|Variable| Coefficient | Standard Error |
+--------+--------------+----------------+
Constant| -26.6786858 2.00029667
LY | 1.62496242 .19517515
LPG | -.05392226 .04216064
LPNC | -.08343220 .17653370
LPUC | -.08467492 .10241732
T | -.01392611 .00477323
The R^{2} for the regression is 0.9649324. The remaining
results in the table are correct except for the standard error given for the ML
estimate of rho. The value given is computed using T-1=51 rather than T=52 in
the denominator. The correct value, to be consistent with (19-32) is 0.02816. The
value of 0.861 for the AR(2) model is obtained as follows: (1) The OLS residuals are
regressed on two lagged values (no constant). Leading observations are filled with zeros.
The two slopes are 0.751941 and -0.022464 as reported. (2) All variables including the
constant term are transformed, then the first two observations are dropped. (3) OLS
using the 50 transformed observations produces the results in the table. (4) The FGLS
coefficient vector is used to recompute the residuals for all 52 observations. (5) The
0.861 is the first autocorrelation coefficient. The new residuals are put in deviation
form - call it du2. The lagged value, du2L is computed and the missing value is filled
with a zero. The autocorrelation coefficient is du2L'du2/du2'du2. In the 7th line of text below the table,
the critical value given for the chi squared distribution, 11.01, should be 11.07.
On page 651, the value 0.78750 given in the third line for the autocorrelation coefficient
was computed using 1-d/2 where d is the Durbin-Watson statistic in (19-23), not using
(19-19) as stated in the text. The motivation for this computation appears at the top
of page 646.
(Professor Randall Campbell, Mississippi State University
and Professor Pedro Bacao, University of Coimbra, Portugal.)
650,651
10/06/07
sample;1-52$
regress;lhs=log_g;rhs=x;ar1;alg=p;maxit=1$ (Prais-Winsten)
crea;et=log_g - x'b$ (Residuals)
crea;ut=et-rho*et[-1]$ Transformed residuals)
crea;ut1=ut[-1]$ Manually fill in the zeros in ut and ut1.
calc;list;ru=ut1'ut/ut'ut;dwru=2*(1-ru)$$
RU = .292161
DWRU = 1.415678
(Professor Pedro Bacao, University of Coimbra, Portugal.)
654
09/28/08
665
03/28/09
679
09/28/08
690
03/28/09
698
3/03/08
709
3/03/08
730
9/28/08
730,731
09/28/08
+--------+--------------+----------------+--------+--------+----------+
|Variable| Coefficient | Standard Error |t-ratio |P[|T|>t]| Mean of X|
+--------+--------------+----------------+--------+--------+----------+
|DY1 | 1.47701*** .11328577 13.038 .0000 -.0152586|
|DY2 | -.51553*** .11276262 -4.572 .0000 -.0099138|
+--------+------------------------------------------------------------+
| Note: ***, **, * = Significance at 1%, 5%, 10% level. |
+---------------------------------------------------------------------+
Time series identification for E
Box-Pierce Statistic = 10.6480 Box-Ljung Statistic = 12.3380
Degrees of freedom = 14 Degrees of freedom = 14
Significance level = .7134 Significance level = .5792
* => |coefficient| > 2/sqrt(N) or > 95% significant.
PACF is computed using Yule-Walker equations.
----+-------------------------------+-------+-------------------------------+
Lag | Autocorrelation Function |Box/Prc| Partial Autocorrelations |
----+------------------+------------+-------+------+------------+-----------+
1 | .063 | |* | .23 | .063 | |* |
2 |-.119 | *| | 1.06 |-.133 | * | |
3 |-.235 | ***| | 4.27 |-.241 | *** | |
4 | .108 | |* | 4.95 | .142 | |** |
5 | .142 | |** | 6.11 | .113 | |* |
6 | .117 | |* | 6.91 | .108 | |* |
7 |-.091 | *| | 7.39 |-.047 | * | |
8 | .058 | |* | 7.58 | .189 | |** |
9 |-.167 | **| | 9.19 |-.321*| **** | |
10 | .034 | |* | 9.25 | .021 | |* |
11 |-.004 | *| | 9.25 | .043 | |* |
12 | .013 | |* | 9.26 |-.072 | * | |
13 |-.134 | *| | 10.31 |-.179 | ** | |
14 |-.076 | *| | 10.65 |-.114 | * | |
----+------------------+------------+-------+------+------------------------+
(Professor Pedro Bacao, University of Coimbra, Portugal.)
745
3/03/08
(2) In the first line of section 22.2.4, "of the of the" should be "of the."
(3) At the end of the page, DF_{t} should be DF_{t}.
(Ryan Polansky, Arizona State University)
751
09/28/08
753
09/22/08
754
10/13/08
755
4/30/08
764
05/19/09
772
3/03/08
774
09/22/08
782
09/22/08
781,782
09/22/08
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|Constant| -7.14055** 3.20817669 -2.226 .0260 |
|GPA | 1.58449* .89781077 1.765 .0776 |
|TUCE | .06023 .12276520 .491 .6237 |
|PSI | 1.61623** .81909913 1.973 .0485 |
+--------+-------------------------------------------------+
Computation of Partial Effects Using the Delta Method
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|Scale | .36209*** .02373008 15.259 .0000 |
|GPA | .57373* .33289382 1.723 .0848 |
|TUCE | .02181 .04393503 .496 .6196 |
|PSI | .58522** .28572006 2.048 .0405 |
+--------+-------------------------------------------------+
(Professor Pedro Bacao, University of Coimbra, Portugal;
recomputed by William Greene, New York University.)
784
09/22/08
785
09/22/08
785
09/22/08
790
09/22/08
790
09/22/08
create;age=wa;age2=wa*wa$
create;kids=(kl6+k618)>0$
create;educ=we$
create;income=faminc/10000$
namelist;x=one,age,age2,income,educ,kids$
probit;quietly;lhs=lfp;rhs=x$
matrix;xb=mean(x)$
create;ab=xbr(age)$
wald;start=b;var=varb;labels=b1,b2,b3,b4,b5,b6
;fn1=n01(xb'b1)*(b2+2*ab*b3)$
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|Fncn(1) | -.00837*** .00282345 -2.966 .0030 |
+--------+-------------------------------------------------+
probit;lhs=lfp;rhs=x;het;hfn=kids,income;mar$
calc;kb=xbr(kids);ib=xbr(income)$
wald;labels=b1,b2,b3,b4,b5,b6,c1,c2;start=b;var=varb
;fn1=n01(xb'b1/exp(c1*kb+c2*ib))*(b2+2*b3*ab)$
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|Fncn(1) | -.00825 .00648692 -1.272 .2035 |
+--------+-------------------------------------------------+
(Professor Pedro Bacao, University of Coimbra, Portugal.)
791
1/30/08
797
3/03/08
802
03/11/08
824
11/06/07
827
12/08/08
844
3/06/10
845
1/13/08
853
07/30/07
853
09/22/08
+--------+--------------+----------------+--------+--------+
|Variable| Coefficient | Standard Error |b/St.Er.|P[|Z|>z]|
+--------+--------------+----------------+--------+--------+
|BG | -.01333*** .00491747 -2.711 .0067 |
|BT | -.13405*** .02569720 -5.216 .0000 |
|CH | -.00108 .01106360 -.097 .9225 |
|AA | 6.59403*** 1.61807045 4.075 .0000 |
|AT | 3.61895*** .83833382 4.317 .0000 |
|AB | 3.32181*** .86915976 3.822 .0001 |
+--------+-------------------------------------------------+
The Log-likelihood (sample shares) for the model with choice based sample weights
should be -218.9929, obtained by fitting the model with only the choice specific
constant terms.
In Table 23.25, in the third row, under Bus, "15(4) should be "15(14),"
(Professor Pedro Bacao, University of Coimbra, Portugal.)
854
10/20/07
877
10/20/07
882
3/03/08
884
10/27/09
888
2/20/08
890
10/17/07
891
3/03/08
898
07/12/08
908
08/07/09
914
08/07/09
915
08/07/09
921
03/11/08
922
3/03/08
959
3/06/10
965
05/13/09
967
3/06/10
968
3/06/10
968
3/06/10
982
1/28/09
997
08/20/07
1055, 1056, 1058
05/08/08
1064
09/22/08
In the expression for the derivatives of the gamma function, lnP should be lnt.
(Professor Pedro Bacao, University of Coimbra, Portugal.)
1076
10/28/08
1076,1077
03/14/10
1081
10/06/07
1084
09/22/08
1084
09/17/07
1086
09/22/08
1086
09/22/08
1087
04/03/08
1088
09/17/07
1092
09/17/07
1107
10/7/09
1123
08/07/09
Hilbe, J., Negative Binomial Regression, Cambridge University
Press, Cambridge, 2007.
(Hai-Anh Dong, The World Bank.)
1133
01/18/08
1133
03/14/09