EasyReg International, 2001 versions
-
Differences between EasyReg 2000
(November 28) and EasyReg International 2001 (April 17):
The main differences are:
-
EasyReg now accept commas and/or dots as decimal delimiters,
regardless the number setting of Windows. Previously you had to change
the number setting of Window, which was a problem on European Windows NT
networks. (This was a major enterprise. All the file operations had to
be modified in order to guarantee correct internal number exchange. Moreover,
the Visual Basic functions Val(), which converts a text string into a number,
and Str$() and Format$(), which convert numbers into text strings, are
sensitive to the number setting of Windows. Therefore, they had to be replaced
with new customized versions that work internally with decimal dots regardless
the number setting of Windows.)
-
Previously the data in folder EASYREG.DAT
was stored in text file INPUT (in default EasyReg data format 1), and text
file DATINFO containing the information about the data type, the frequency
and date of the first observation in the case of time series, and the total
number of observations. Now the data is stored in two random access files,
INPUT1.RAN and INPUT2.RAN, so that the data can be imported much faster
in EasyReg.
-
If you start-up EasyReg International in a
folder containing a sub-folder EASYREG.DAT created by EasyReg 2000 or older,
the files INPUT and DATINFO are automatically converted into the random
access files INPUT1.RAN and INPUT2.RAN.
-
The Johansen cointegration approach now includes innovation
response analysis with cointegrated vector error correction models.
-
Two-dimensional nonparametric SMINK density estimation
(was one-dimensional only).
-
The 3D plots have been upgraded. My home made 3D plot
procedure now generates (almost) professional looking pictures.
-
The teaching tools menu has been augmented with a module
which plots of the bi-variate normal distribution, in order to demonstrate
how a near-singular multivariate normal distribution looks like.
-
EasyReg now accepts Excel files in CSV format directly.
The CSV format is now the default data format, but the former EasyReg default
data file format (format 1) is still supported. Therefore, it is convenient
but not necessary to have Microsoft Excel installed.
-
Now you choose the input file first, and then the location
of the folder EASYREG.DAT, rather then the other way around in the previous
version. The latter caused too much confusion.
-
The GMM module now comes with a guided tour.
-
All the EasyReg modules except the start up module EASYREG
and EASYREGN
are now Multiple Document Interface (MDI) programs, so that it is now safe
to minimize EasyReg while it is doing number crunching in the background.
Previously EasyReg could then crash. Moreover, the windows of EasyReg 2000 do
not stick to the background window, and could therefore shift or be covered
up by other EasyReg windows. In that case you had to click the background
window in order to restore them. In EasyReg International the background
windows have been replaced with MDI windows, with sticky child windows:
the child windows cannot be moved, and in the case that two child windows
are visible the smaller one will get and keep the focus.
-
The module EASYREGN now allows network administrators
to initialize EasyReg for network use.
-
The nonlinear regression module is not yet available
in EasyReg International, but will be in the next release.
-
plus many other upgrades and changes, too many to list
here (as users of EasyReg 2000 will find out).
-
Differences between the versions April 17 and May 1, 2001 :
- The number of degrees of freedom of the Wald test in the OLS module was incorrect if a subset of observations was selected. This bug has been corrected.
- If the number of selected observations was less than 30, the modules for OLS regression, Tobit analysis, limited dependent variables regression, quantile regression, and two-stage least squares, refused to continue. This bound has been removed and replaced by a warning.
- CSV files containing data on only one variable were not accepted. They are now.
-
Differences between the versions May 1 and May 2, 2001 :
- Missing values at the end of a record in CSV files were not recognized. They are now.
-
Differences between the versions May 2 and June 25, 2001 :
- The version date in the title bar of the EasyReg main window
was intended to be the compilation date, but turned out to be the installation date.
Therefore, it was not clear after installation whether EasyReg needed to be upgraded.
This bug has been corrected.
- If the EasyReg folder is write-protected, module EASYREG will
now call EASYREGN, and close itself, rather than giving an error message.
- A non-essential bug in the module for the ICM test has been corrected:
The displayed list of instrumental variables was not always correct.
- The ARCH test was incorrect if the model contains lagged variables and/or
a subset of observations is used.
This bug was inherited from EasyReg 2000 and previous versions, and corrected in both
EasyReg 2000 (16 bit and 32 bit versions),
and EasyReg International 2001.
- The 2-dimensional plots in the modules TSPLOT (time series plot),
KDENSITY (kernel density), KERNELREG (kernel regression), and SCATTER
(scatter diagram) are now automatically resized if you run EasyReg in full screen mode.
- Some cosmetic upgrades in the module CUNITR of Bierens' complex unit root test.
-
Differences between the versions June 25 and June 30, 2001:
- In the case of dat your data contains huge
numbers, the single equation modules and the GMM module may give an overflow error message.
The codes of these modules have been adjusted so that this problem no longer occurs.
(Thanks to Dr. Louise C. Keely
for pointing this problem out to me.)
- In the case that missing values in cross-section data are scattered around rather than at
the beginning and the end of the data file only, you could get an error message saying:
"Error in subroutine MakeOBSTMP". This was due to
a bug, and has been repaired in all the single equation modules.
- The results of the nonlinear nonparametric co-trending analysis (module COTREND) in
EasyReg International were different from the corresponding results in EasyReg 2000. This was
due to bugs in EasyReg International. Thanks to Marianna Belloc for pointing this problem out to me.
I have also checked the cointegration tests
(Johansen, module COINTJ, and Bierens, module COINTB), and the two versions of EasyReg
gave the same results. However, the check box
"Write output to file EASYREG.DAT\OUTPUT.TXT when done" in these
three modules have now been set to default value "checked."
-
Differences between the versions June 30 and August 12, 2001:
-
A bug in the Chebishev time polynomials part
of the OLS module has been repaired.
-
The Kiefer-Vogelsang-Bunzel (KVB)
test of linear parameter restrictions in linear regression models has been
included in the "What to do next?" module.
Reference:
- Kiefer, N. M. , T. J. Vogelsang, and H. Bunzel (2000): "Simple Robust
Testing of Regression Hypotheses", Econometrica 68, 695-714.
-
The nonlinear least squares module
(NLLS) and a guided tour on how to estimate a CES production function by NLLS
has been included.
- User-defined maximum likelihood estimation
(module ML) together with a guided tour
has been included.
- The Help and About menu items are now displayed
via the new module ABOUT. In due course this
module will be replaced by help files which are under construction by
Dr. Guillermo Atilano at UAM, Mexico.
- A new module, TOURS, has been added, which gives direct access to the available guided tours. This module
is activated via the new menu item "Tours" in the EasyReg main window.
-
Differences between the versions August 12 and September 12, 2001:
- User-defined nonlinear General Method of Moments (module NLGMM)
has been added, including a guided tour.
- The three user-defined nonlinear estimation modules (NLLS, ML, and NLGMM)
have been endowed with the following new transformations:
- Subtract
- Ratio
- Normal(z) (Density of the N(0,1) distribution)
- Dummy z = 0
- Dummy z > c, with c a constant
- Dummy z = c, with c a constant
- Dummy z < c, with c a constant
- Integral of F(a+b.x)f(x)dx
- Integral of I(x > c)F(a+b.x)f(x)dx
- Integral of I(x < c)F(a+b.x)f(x)dx
where f(x) and F(x) are the density and distribution function, respectively,
of the standard normal distribution, and I(.) is the indicator function.
The integrals may occur in simultaneous limited
dependent variables models. Also, a new window has been added to these
modules explaining the less obvious transformations.
Moreover, error messages have been added
for functions with a restricted argument, for example "LOG(z)" and "SQR(z)", and
transformations with the potential for an overflow error, such as "Reciprocal: 1/z".
Furthermore, the constant 1 is automatically added to the list of ML, NLLS and NLGMM
model variables.
- If you add a constant 1 or time trend or seasonal dummies to the data set and select
them as X variables, you will no longer be able to choose them separately as deterministic
X variables. Moreover, the option of chosing Chebishev time polynomials as deterministic
X variables in an OLS regression will then be blocked.
- The menu items in the "What to do next?" module (NEXTMENU) which are not applicable
are now hidden. Previously they were only disabled.
- The module for deleting variables from the input file (DELVAR) did not allow to
delete variables if the chosen variables had no valid observations in common. It does now.
- The "About..." menu items in the EasyReg main module for which the answers can be found
on the EasyReg web page have been deleted. The remaining "About..."
menu items, together with the items in the former "Help!" menu, have been re-arranged in a
single Help menu.
- A bug in module PICVIEW for viewing and editing picture files has been
repaired. If the path of the start-up folder contains blanks and
you choose to import a picture file in MS Paint, the path of the picture file involved was
cut off at the first blank, causing an error.
- The limited dependent variables module LIMDEP has been augmented with ordered probit
and logit.
- Note that the new or upgraded guided tours are no longer included in the
upgrade files. You have to download them seperately.
-
Differences between the versions September 12 and October 5, 2001:
- The two-stage least squares module (TSLS) was not quite "International",
so that it did not accept commas as decimal delimiters.
It does now. Thanks to Javier Rojas for pointing the problem out to me.
- The auto- and cross-correlation module (AUTOCOR) refused to work
if the effective length of the time series is less than 60.
This was a deliberate threshold,
because for such a small time series the (asymptotic) results are unreliable.
Nevertheless, upon request I have lowered the threshold to time series
of effective length 15.
- It appears that the fractiles of the T-Tilde test in Table 4 in the paper:
- Bierens, H.J. (1997), "Testing the Unit Root with Drift Hypothesis
Against Nonlinear Trend Stationarity, with an Application to the U.S.
Price Level and Interest Rate", Journal of Econometrics 81, 29-64,
are incorrect. Click here for the corrected Table 4.
These fractiles have been corrected in the unit root tests module (UNITROOT).
Thanks to Jorge Belaire-Franch for pointing this problem out to me.
- If you conduct Johansen's cointegration analysis (module COINTJ) with the
options
"Cointegrating restrictions on the intercept parameters imposed", and p = 1,
you get an error message "Subscipt out of range". This was due to a bug and
has been repaired in EasyReg International 2001 as well as in the 16 and 32
bit versions of EasyReg 2000.
Thanks to Matthew Yeung for pointing this problem out to me.
- A new menu item "Number setting of Windows" has been added to the File menu
on the EasyReg main window.
- The instructions on how to arrange panel data files has been improved,
in module ABOUT and the guided tour on linear GMM.
-
Differences between the versions October 5 and November 20, 2001:
- A non-essential bug in the OLS module disabled the extension of a time series regression to
linear time series models with ARCH and GARCH errors. This has been repaired.
- Due to a bug, the OLS forecast standard errors were based on the heteroskedasticity consistent variance
matrix rather than on the standard variance matrix.
This bug has been repaired. Thanks to Professor Nick Robinson (North Island College)
for pointing this problem out to me.
- A new transformation, the fixed point solution
- z > 0: z^z1 = z2.z + z3, z1 > 0,
has been added to the modules NLLS, ML and NLGMM. I needed it for my research.
- The appearance of EasyReg onder Windows 2000 has been improved.
- An example of an EasyReg advertisement has been added to the "What to do next?"
module (NEXTMENU).
- The user-defined nonlinear models modules have been arranged in a separate
sub-menu of the "Menu" item on the EasyReg main window.
- The module DATACSV for importing CSV input files has been improved.
- The module for tabulating data (DATATAB) did not work if the chosen
variables had no valid observations in common. It does now.
- A new guided tour on OLS estimation has been added.
- The corrected Table 4 with the quantiles of the
T-Tilde test in the paper:
- Bierens, H.J. (1997), "Testing the Unit Root with Drift Hypothesis
Against Nonlinear Trend Stationarity, with an Application to the U.S.
Price Level and Interest Rate", Journal of Econometrics 81, 29-64,
was still incorrect, as was pointed out to me by Jorge Belaire-Franch.
Therefore, I have reprogrammed the Monte Carlo simulation program from scratch, and
recomputed the quantiles involved. These quantiles are now in
accordance with the results computed by Jorge Belaire-Franch,
and have been corrected once more in the unit root tests module (UNITROOT).
-
Differences between the versions November 20 and December 31, 2001:
- A non-essential bug in the Bierens cointegration module (COINTB) caused occasionally
the cointegrating vectors to be displayed in a different order than the order of the
generalized Eigenvalues involved. This bug has been repaired. Thanks to Dave Cushman for
pointing this problem out to me.
- A few non-essential bugs have been repaired.
- Various cosmetic upgrades.
Back to the EasyReg page
Current version of EasyReg