1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319

Changes in Version 1.28
o Some examples from the AER book lead to small numeric differences across different
platforms. Hence, these have been excluded now from tests/Ch*.R with an IGNORE_RDIFF
comment for CRAN.
o The examples on the manual pages for the books CameronTrivedi1998, Greene2003,
WinkelmannBoes2009 are now excluded from testing to reduce the computational demands
on CRAN. The corresponding analyses are mostly available on the manual pages for the
underlying data sets, though.
Changes in Version 1.28
o Starting from survival >= 3.16 the survival provides (or corrected) some standard
S3 methods for "survreg" objects: fitted(), nobs(), weights(), vcov(). Previously,
these were registered by AER" in order to work for "tobit" objects. For now,
AER registers the methods for "tobit" objects. In the future, the methods will
be dropped by AER altogether and inherited from survival (when AER depends on
survival >= 3.16).
Changes in Version 1.27
o Diagnostic tests ivdiag() for _weighted_ instrumental variables regression
was incorrect as weights were erroneously ignored (detected and tested by
Jonathan Siverskog).
o vcov() method for "survreg" objects in "survival" currently (version 2.441.1)
provides no row/column names. Hence, a new vcov() method for "tobit" objects
was added in "AER" mimicking the naming conventions from survival:::summary.survreg.
Analogously, a bread() method was added for "tobit" objects that calls
the vcov() method internally.
o linearHypothesis() method for tobit() objects no longer assumes that a scale
parameter was estimated as part of the model. This enables the somewhat
unusual but possible case of summary(tobit(..., dist = "exponential")).
o In examples/demos/tests with calls to set.seed(...) the RNG version is now fixed
to suppressWarnings(RNGversion("3.5.0")) to keep results exactly reproducible
after fixing RNG problems in R 3.6.0.
o New errata item in vignette("AER", package = "AER"): The formula interface for
pgmm() has changed. As of "plm" version 1.70, the function dynformula() is
deprecated. Examples, tests, and demos have been adapted accordingly.
o In tests/ChIntro.R some quantreg computations are now wrapped into try()
because the summary method yields nonfinite standard errors on some platforms.
o In ?CameronTrivedi1998 one version of the negbinnegbin hurdle model is now
wrapped into try() because evaluating the loglikelihood at the given start
values becomes too unstable on some platforms.
Changes in Version 1.26
o Added update() method for "ivreg" objects that correctly handles the
twopart righthand side of the formula (suggested by Matthieu Stigler).
o Use pdata.frame() instead of plm.data() as preparation for modeling
with "plm" and certain "systemfit" functions (requires systemfit
1.120).
o The model.matrix() method for "ivreg" objects erroneously dropped
the dimension of singlecolumn projected regressor matrices. This
propagated to the estfun() method and hence lead to problems with
sandwich covariances (reported by Justus Winkelmann).
o In a bug fix the survival package changed the internal structure of
survreg()$y starting from survival 2.427. This leads to incorrect
summary() output for "tobit" objects which has been worked around
now. Thanks to Terry Therneau for pointing out the problem and suggesting
a fix.
Changes in Version 1.25
o Support for aliased coefficients in ivreg() (suggested and tested by
Liviu Andronic).
o New data sets GoldSilver, MotorCycles2 and MSCISwitzerland, taken from
Franses, van Dijk, Opschoor (2014): "Time Series Models for Business and
Economic Forecasting", 2nd ed. For replication of the corresponding
examples, several packages were added to the list of 'suggested' packages
(including fGarch, forecast, longmemo, rugarch, vars).
o Small improvements in DESCRIPTION/Imports and NAMESPACE for R CMD check.
Changes in Version 1.24
o Reference output updated for recent versions of R.
Changes in Version 1.23
o Package "splines" is loaded explicitly if needed (rather than assuming
that it is loaded along with "survival").
o Some URLs in the manual pages had been outdated and are updated (or omitted)
now.
Changes in Version 1.22
o Another bug fix in the new summary(ivreg_object, diagnostics = TRUE). If
sandwich standard errors (or other vcov) were used, the chisquared form
rather than the F form of the diagnostic Wald tests was computed and hence
the pvalues were incorrect.
o If there is more than one endogenous variable, summary(ivreg_object,
diagnostics = TRUE) now reports separate tests of weak instruments for
each endogenous variable.
Changes in Version 1.21
o Bug fix in the new summary(ivreg_object, diagnostics = TRUE). If there
is more than one endogenous variable, the degrees of freedom (and hence
the associated pvalues) for the Sargan test were too large.
o The examples employing "rgl" for 3d visualizations (e.g., for the SIC33
data) are not tested anymore in R CMD check (as "rgl" currently has some
problems on CRAN's checks for OS X).
Changes in Version 1.20
o The summary() method for ivreg() now has a diagnostics=FALSE argument.
If set to TRUE, three diagnostic tests are performed: an F test of
the first stage regression for weak instruments, a WuHausman test
for endogeneity, and a Sargan test of overidentifying restrictions
(only if there are more instruments than regressors).
o Added new data set EquationCitations provided by Fawcett & Higginson
(2012, PNAS).
o Changes in Depends/Imports/Suggests due to new CRAN check requirements.
In particular, the "Formula" package is now only imported (but not
loaded into the search path).
Changes in Version 1.19
o Recompressed data sets in package to reduce file storage requirements.
o ivreg() failed when used without instruments. Now fixed.
o The summary() for ivreg() displayed the degrees of freedom of the
overall Wald test incorrectly (although the pvalue was computed
correctly).
o Some technical changes for new R 2.14.0, e.g., adding Authors@R in
DESCRIPTION, recompressing data, etc.
Changes in Version 1.18
o The hat values for instrumental variables regressions are now
computed in the hatvalues() method and not within ivreg.fit()
to save computation time for large data sets.
o Added nobs() method for "survreg" objects (and thus "tobit" objects).
Modified "ivreg" objects so that default nobs() methods works.
o Labeling in coeftest() method for "multinom" objects with
binary responses has been fixed.
o Example 21.4 in ?Greene2003 now employs the scaled regressor
fincome/10000.
Changes in Version 1.17
o Adapted some example in ?Greene2003 in order to work both with
the old and new "dynlm" package. dynlm() now provides convenient
support for linear time trends via dynlm(y ~ trend(y)) etc.
Changes in Version 1.16
o Adapted code/examples/tests to new car version which has deprecated
linear.hypothesis() in favor of linearHypothesis().
Changes in Version 1.15
o CPS1985 now has 534 observations (not 533 as in prior releases),
the original observation 1 had been omitted inadvertently. See
also the errata in vignette("AER", package = "AER").
o Data and examples for Winkelmann and Boes (2009),
"Analysis of Microdata" (2nd ed.) have been added. For details and
extensive (but not quite complete) replication code see
help("WinkelmannBoes2009") as well as help("GSS7402") and
help("GSOEP9402").
o As announced in the changes for version 1.10 of the "AER" package,
the variable "earnings" has now been removed from the
PSID1976 (aka Mroz) data. In 1.10 it was renamed to "wage"
to avoid confusion with other data sets.
o The coeftest() method for "polr" objects used to return wrong
standard errors (and hence wrong z tests) for the last intercept.
This was caused by an inconsistency between the summary() and
vcov() methods for "polr" objects which has been improved in
recent versions of the "MASS" package. The correct results are
computed by coeftest() for "polr" objects computed with MASS
version >= 7.36. See also the errata in vignette("AER",
package = "AER")
o The paper describing the various versions of the Grunfeld data
has been accepted for publication in the German Economic Review.
An updated version of the manuscript and associated replication
files  mostly based on data("Grunfeld", package = "AER") 
is available from http://statmath.wu.ac.at/~zeileis/grunfeld/.
o Added lrtest() method for "fitdistr" objects with intelligible
model name (instead of the usual formula for formulabased models).
Changes in Version 1.14
o ivreg() now uses the "Formula" package (>= 0.20) for processing
of its model formulas. However, this only affects the internal
computations, the user interface has remained unchanged.
o Numerous spelling improvements in the documentation (thanks to
the new aspell() functionality in base R).
Changes in Version 1.13
o Added PSID7682 data set which contains the full Cornwell & Rupert
(1988) panel data for the years 19761982. This should be used
for estimation of the HausmanTaylor model in Exercise 6 from
Chapter 3 (instead of PSID1982 which does not provide panel data
but only the crosssection for 1982). See the errata and the
manual page for more details.
o Fixed overall Wald test in summary() for "tobit" models
with intercept only.
Changes in Version 1.12
o New errata item in vignette("AER", package = "AER"): The comment
regarding the output from the Johansen test (p. 169) is in error.
The null hypothesis of no cointegration is not rejected at the
10% level (only at 15% level).
o Enhancements of the CigarettesSW examples from Stock & Watson.
o Fixed overall Wald test in summary() for "tobit" models
without intercept.
o Improved "rgl" code in the SIC33 example.
o The variable "gender" in the Parade2005 data set was
wrong for observation 70. It is now "male" (not "female").
Changes in Version 1.11
o A new improved version of the "plm" package is available
from CRAN (version 1.11). This fixes a bug in the summary()
of "plm" objects, see the vignette/errata for details.
Furthermore, there is now a vcovHC() method for "panelmodel"
objects: It gives equivalent results to pvcovHC() but
is now the recommended user interface and hence used
in the examples of some manual pages (see e.g. ?Fatalities).
Changes in Version 1.10
o Some variable names in the PSID1976 (aka Mroz) data
have been renamed: "earnings" is now called "wage"
(to avoid confusion with other data sets), the previous
variable "wage" is renamed as "repwage" (reported wage).
Currently, "earnings" is kept; it will be removed in future
releases.
o Documentation of the Grunfeld data has been enhanced and
updated. Much more details are available in a recent
technical report: Kleiber and Zeileis (2008), "The
Grunfeld Data at 50", available from http://epub.wuwien.ac.at/.
o Multinomial logit examples using Yves Croissant's "mlogit"
package have been added for the TravelMode and BankWages
data sets.
o Vignette/errata updated.
Changes in Version 1.01
o Small changes for R 2.8.0.
Changes in Version 1.00
o official version accompanying the release of the
book (contains all code from the book in demos
and tests)
o See the new
vignette("AER", package = "AER")
for an overview of the package and a list of errata.
Changes in Version 0.90
o release of the version used for compiling the final
version of the book for Springer
Changes in Version 0.20
o first CRAN release of the AER package
