EasyReg International, 2004 versions
 Differences between the versions
October 30, 2003 and March 1, 2004:
 Bug repairs
 The option to write the discrete dependent variable probabilities
to the input file did not work correctly for ordered probit and logit,
because the probabilities did not add up to 1.
They do now. Thanks to Peter Jamieson for pointing this problem out to me.
 Also, a related bug in the likelihood functions of the ordered probit and logit
models has been repaired. However, this bug did not affect
the ML estimates and the tvalues.
 The guided
tour on discrete dependent variable models has been updated accordingly.
 In the October 30, 2003, version a new module, ERIUPGRADE.EXE, was added.
This module checks whether upgrades of EasyReg modules
(EXE files) and auxiliary files (such as PDF files, but not the guided tours
and the database files) are available, and if so allows you to download these
files directly from the web as selfextracting WinZip files. However, it is
possible that some of these selfextracting WinZip files get damaged in the
downloading process.
Therefore, the upgrade procedure has been modified as follows.
If newer EasyReg files are available a web page "Upgrade EasyReg International"
will be created and launched with further instructions how to download these files.
At the same time a program COPYNEWER will be launched (in minimized form) which
copies the newer files to the EasyReg folder (upon confirmation) when the
downloading is successfully completed.
 If you conduct VAR innovation response
analysis or Johansen's cointegration analysis and your time series are short,
you may get an error message "Runtime error 5". This happens if the VAR or
VECM order p is chosen too large. Now you will get a warning message if you
choose p too large.
 Occasionally the regression line in the scatter diagram (module SCATTER.EXE)
is cut off at the horizontal borders. Now the regression line stays always
between the horizontal borders.
 If the EasyReg folder is writeprotected and EASYREG.EXE is run then
EasyReg should start up automatically with EASYREGN.EXE. This did not always
work correctly. It does now.
 Additions and improvements
 The modules that use the Akaike, HannanQuinn and Schwarz information
criteria now contain a help button or a menu item which explains these
criteria, via a new PDF file INFCRIT.PDF.
 When you run the discrete dependent variable module (LIMDEP) you will now
have the option (in the "What to do next?" module) to write the probabilities
involved (in sample as well as outof sample) to the input file. You can
view these probabilities via Menu > Data analysis > Data table.
 The presentation of the OLS estimation results (module OLS.EXE) has been
improved. In particular, the standard errors of the OLS estimates are now
presented as well. The reason for this update is to make the EasyReg output
more in tune with
 Stock, James H., and Mark W. Watson (2003),
Introduction to Econometrics,
which I am using to teach undergraduate econometrics
(ECON 490)
 It is now possible to upgrade the guided tours in the same
way as the EasyReg system files, via menu "WWW > Upgrade guided tours".
If newer guided tours are available a web page "Upgrade EasyReg International
guided tours"
will be created and launched by program ERITOURSUPGRADE with further
instructions how to download these guided tours.
At the same time a program COPYNEWERTOURS will be launched (in minimized form) which
copies the newer guided tours to the EasyReg folder (upon confirmation) when the
downloading is successfully completed. Note that this option is only available
if the EasyReg folder is not writeprotected.
 The explanation of how the network version of EasyReg works (in module
EASYREGN.EXE via the "Tell me more!" button) has been improved.

Differences between the versions
March 1, 2004 and October 6, 2004:
 Bug repairs
 Upon installation of EasyReg on a network, the administrator is asked to provide
the path to the Microsoft Excel executable file EXCEL.EXE on the network if the network
PC's do not have standalone versions of Excel. If so, the path to the network version
of Excel is stored in the EasyReg network folder as file EXCEL.TXT. However, this
information was never passed on to the users. Now it is, in the following way. If you
open "File > Find and/or open Microsoft Excel" to activate EasyReg module FINDEXCEL,
and the path to Excel is not yet stored in the user's EasyReg settings folder,
FINDEXCEL will now copy the file EXCEL.TXT from the EasyReg network folder to the
user's EasyReg settings folder.
 When you run an OLS regression using a subsample of crosssection data, the first
and last observation numbers of the subsample were not written to the output file.
 The matrix inversion procedure in EasyReg did not check correctly for singularity.
As a consequence, the OLS module detected perfect multicollinearity at too late a stage,
if variance estimators turn out to be zero or negative.
Thanks to Nick Kiefer for pointing this problem out to me.
Various safeguards have now been
build in to prevent a singular matrix to be inverted. In particular, the smallest
eigenvalue of the matrix (1/n)X'X is calculated before this matrix is inverted.
If this eigenvalue is zero you will get a message that you have
to respecify your model, and you cannot continue. If this eigenvalue is positive but less
than 0.00001 you will now get a warning message. Moreover, the matrix inversion is now done
in the highest possible precision (previously only double precision was used).
These changes have been implemented in all other modules as well.
 If the module for upgrading the guided tours (ERITOURSUPGRADE) reports that the
guided tours are uptodate, and you doubleclick the list box, you got an "Error 5"
message, and EasyReg crashed.
 In the VAR innovation response analysis module (VAR) you have the option
to write the numerical values of the innovation responses and their standard
errors to the output file OUTPUT.TXT. However, if you get the message that the
text box is full and you check "Store output" then only the incomplete results
were written to the output file. Now the complete results will be written to the
output file. Thanks to David Porupka for pointing this problem out to me.
 If you conduct the Johansen cointegration analysis using seasonal data,
and you include seasonal dummy variables in the VECM, the HannanQuinn and
Schwarz information criteria for selecting
the VECM order p did not work, due to a bug.
 It appears that occasionally the input files get damaged when new variables are written
to these files. This may occur if the data set is large, and you accidently
click a control which starts a reading action from the input files while the writing process
is not yet finished. To prevent this, the EasyReg windows involved are now disabled while data is
written to the input files, and enabled when the writing process is finished.
 Additions and improvements
 The heteroskedasticity consistent variance matrix, tvalues and standard errors
in the OLS module are now corrected for the degrees of freedom. The reason for this
update is that the same is done in
 Stock, James H., and Mark W. Watson (2003), Introduction to Econometrics,
which I have used to teach undergraduate econometrics (ECON 490).
 The module PVAR which computes the critical values and distribution functions
values of various continuous distributions (Tools > Distribution tools >
Continuous distributions) now enables you to generate a table
of the distribution function values of the standard normal distribution, and tables
of critical values of the chisquare, t and F distribution.
 The OLS module now provides the option to display the estimation results in
floating point format.
 If you invert a matrix using the matrix calculator (under Tools > Matrix tools)
you will now have the option to see stepbystep how the GaussJordan iterations for
inverting a matrix works, and how the determinant of the matrix involved is calculated.
 The appearance of EasyReg under Windows XP has been improved,
by replacing the big and ugly title bars of the child windows with smaller
title bars.

Differences between the versions
October 6, 2004, and November 17, 2004:
 Bug repairs
 The structural part of the VAR innovation response module did not
always work correctly. The problem was due to the way this module solved the
structural equations. Therefore, the solution algorithm has been redesigned
and reprogrammed to make it more robust. If the system cannot be solved,
EasyReg will now display all the solution steps up to the point where it
got stuck. Only structural systems that can be solved analytically will be
accepted.
 If you change the level from Research level to a lower level you got
an error message, due to a recent bug.
 There was a nonessential bug in the ADF unit root test: In the case
of ADF test 2 the correct hypotheses involved should be
"H0: Unit root; H1: Stationarity around a constant" rather than
"H0: Unit root with drift; H1: Linear trend stationarity". Thanks to
Emin Karagozoglu for detecting this bug.
 Additions and improvements
 The time series plot module now also displays the value of the time series
in the title bar if you move the mouse pointer over the plot.
 Some nonessential cosmetic upgrades in the data import and transformation
modules.
 A new module, SURVIVAL, has been added. This module estimates a proportional
hazard model without unobserved heterogeneity, where the duration involved
is only observed within brackets. This module is available via
"Menu > Single equation models > Proportional hazard model", for crosssection data
and advanced or research level only. A guided tour is under construction and
will be available soon.
 The procedure for computing the determinant of a matrix has been improved,
to make it more accurate in the nearsingular case.