EasyReg: 2000 versions

• Difference between Version 1.30 (October 1999) and Version 2000 (January 2000):
• A new guided tour (modules TNLLS.EXE and TNLLS2.EXE) has been added which explains step-by-step how to estimate a CES production function by nonlinear least squares. Also, the list of transformations has been augmented with four new transformations:
• LOG(1+z)/z
• (EXP(z)-1)/z
• SIN(z)/z
• (COS(z)-1)/z
These additions are due to a query of an EasyReg user.
• The EasyReg database can now also be viewed if you have not yet imported data in EasyReg.
• Some cosmetic upgrades.
• Difference between Version 2000 (January 2000) and Version 2000 (February 1):
• The main difference is in the 32-bit version:
• The January version was only partially a 32-bit program, but the February version is a genuine 32-bit program.
• The access to the internet now goes directly through the default web browser.
• The WWW menu has been extended with links to my email address and Peter Sephton's review of EasyReg 1.12 in the Journal of Applied Econometrics.
• The data files format menu items have been moved from the 'About' menu to the 'File' menu.
• The Bierens-Guo stationarity tests 2-4 in EasyReg are conducted differently from the proposals in the published paper. EasyReg now explains why (16-bit and 32-bit versions).
• Difference between the February 1 and February 7 versions:
• There was an error in my working paper "Complex Unit Roots and Business Cycles: Are They Real". In particular, the claim that under stationarity the ordinates of the standardized periodogram are bounded from above by chi-square (2) variates is in general not correct. This error affected two modules:
• Periodogram (PERGRAM.EXE): The confidence bands under option "Continuous 2" were only correct if the time series involved is white noise. This option has therefore been deleted.
• Test for complex unit roots (CUNITR.EXE): The size of the test of the stationarity hypothesis against the complex unit root hypothesis was incorrect. This stationarity test has therefore been deleted. The test of the complex unit root hypothesis against stationarity is correct, though, and is still available in EasyReg.
The updated version of EasyReg has been put on the web on February 8, 2000, at 1:30 PM Eastern US time.
• Difference between the February 7 and February 13 versions:
• A new module (ARIMA.EXE) has been added, which conducts ARIMA modelling, estimation and forecasting. I needed this for my course ECON 511 (Time Series Analysis).
• The modules for changing the fonts have been speeded up.
• There was a bug in the module for transforming variables. The time series transformation "Partial sum" did not handle missing values correctly. It does now.
• There was a bug in the module for nonlinear cotrending analysis: If you choose another value of the parameter "alpha" than the default value 0.5, the test for linear restriction on the cotrending vectors still used 0.5. This has been corrected (The error does not affect the results in the paper). Thanks to Dave Cushman for pointing this out to me.
• Since it is too time consuming to download EasyReg completely, in particular the 32-bit version, these upgrades are now in the form of a separate self-extracting Winzip file, ER32UPGR.ZIP (2.6 MB). If you have already installed one of the February 2000 32-bit versions of EasyReg, there is longer a need to re-install EasyReg completely.
• Difference between the February 13 and and February 14 versions:
• The day after I had put the EasyReg version of February 13, 2000, on the web, I found a non-essential but annoying bug in module DATA.EXE of the 32-bit version, due to a conversion problem. Effective February 14, 2000, this has been repaired. The upgraded module DATA.EXE has been included in the upgrade file ER32UPGR.ZIP.
• Difference between the February 14 and February 17 versions:
• There were quite a few bugs in the modules for ARMA errors, GARCH errors, and ARIMA forecasting. In particular, the t-values were a factor 10 too high, and the computation of the residual sum of squares in linear and nonlinear models with ARMA errors was incorrect for ARMA(p,q) errors with p > 1. The same applied to the ARIMA estimation and forecasting module and the GARCH module. These bugs have been repaired effective February 17, 2000. Thanks to Marius Ooms for pointing them out to me.
• Difference between the February 17 and March 14 versions:
• The 16-bit version of Easyreg has been taken off the web.
• The computation of the partial autocorrelation function took too much time. It is now possible to specify the maximum number of lags in order to reduce the computation time.
• The search for wordprocessors in order to view the output file did not work in the 32-bit version of EasyReg. It does now.
• The module for testing complex unit roots has been upgraded and extended (nothing was wrong with the previous one, though).
• If you download an input text file from the internet, the line breaks were not recognized by EasyReg. It does now.
• The partial autocorrelation function was incorrect. EasyReg computed the m-th partial autocorrelation of a time series Y(t) as the correlation between Y(t-m) and the residual of the regression of Y(t) on Y(t-1),....,Y(t-m+1) (which I have seen somewhere in the literature). However, most authors define the m-th partial autocorrelation as the coefficient of Y(t-m) in the regression of Y(t) on Y(t-1),....,Y(t-m). EasyReg has now adopted the latter. (Thanks to Juan Molina Capilla for pointing this problem out to me.)
• Difference between the March 14 and March 20 versions:
• The disabeling of menu items in the absence of a data file did not work correctly. It does now.
• The ARIMA menu item did not get disabled in the case of cross section data. It does now.
• A new module, EXCEL.EXE, had been added. This module converts an Excel file saved as a CSV (comma delimited) file to an EasyReg data file in default format.
• Difference between the March 20 and April 3 versions:
• The structural part of the VAR module in EasyReg had a minor bug, which occasionally caused an "end of file" error.
• The option of conducting VAR innovation response analysis without confidence bands has been deleted, because it served no purpose.
• The Integrated Conditional Moment (ICM) test for median regression models has been extended to general quantile regression models.

See:
• Bierens, H.J., and D. Ginther (2000): "Integrated Conditional Moment Testing of Quantile Regression Models", under revision for the special issue of Empirical Economics on Economic Applications of Quantile Regression.
• The guided tour for first-time EasyReg users on how to import data (formerly TOUR.EXE) has been replaced by a web based tour (starting at DATA.HTM), which uses the default web browser off-line. The reason for this replacement is twofold:
• The file TOUR.EXE was very big. The web browser version is much smaller in size.
• The new version is easier to upgrade, hence it is now more detailed.
• The module for simulating the null distributions of unit root tests has been extended with the option of wild bootstrapping. Previously, the simulation was done by replacing the error of the AR model onder the null hypothesis by random drawings from the normal distribution with zero mean and variance the estimated variance of the errors. This option is still available, but is no longer the default. Wild bootstrapping takes possible (autoregressive) heteroskedasticity into account, by replacing each error by a random drawing from the normal distribution with zero mean and variance the squared OLS residual. Wild bootstrapping is now the default option.
• The module for nonlinear least squares now supports templates. Also, a few model specification correction buttons have been added, so that you can undo the last transformation, or the last selection of variables, when you build-up the nonlinear model.
• The module "Teaching tools" has been augmented with a window which demonstrates the shapes of various Wiener processes (Standard Wiener process, Brownian bridge, demeaned Wiener process, and detrended Wiener process).
• Most of the modules which have been upgraded since February 1, 2000, have been "cosmetically" upgraded as well, in the sense that the controls now have a three-dimensional look. The latter might cause the EasyReg upgrade not to work properly (I cannot test it, because my systems are already configurated for this look). If so, please reinstall EasyReg before you email me.
• Difference between the April 3 and April 9 versions:
• The partial autocorrelation function (in module AUTOCOR.EXE) is now endowed with standard errors, and the minimum horizon has been made the default. These changes are due to suggestions of my teaching assistant, Sergei Izmalkov.
• When you double click the plots of the auto- and cross-correlation functions (in module AUTOCOR.EXE), and the VAR innovation responses (in module VAR.EXE), the printed horizon will now be lined up correctly.
• If you convert a large Excel worksheet in CSV format to an EasyReg data file (module EXCEL.EXE), the text box in which the first 50 records of the CSV file are displayed may reach it maximum capacity. Previously, this would cause a fatal error. Now you will only get a warning message.
• A guided tour has been added to module EXCEL.EXE. This guided tour, which uses the default web browser, explains in detail how to convert an Excel worksheet saved in CSV format to a data file in EasyReg default format. The guided tour can be accessed via the menu item 'Guided tour' on the background window of module EXCEL.EXE.
• The guided tour for first-time EasyReg users is now also accessible via the menu item 'Guided tour' on the background window of module DATA.EXE.
• The guided tour on nonlinear least squares estimation (files TNLLS.EXE and TNLLS2.EXE) has been replaced with a new version (starting at NLLS.HTM) which uses the default web browser. This guided tour is now also accessible via the menu item 'Guided tour' on the background window of module NLLS.EXE. Moreover, when you start-up NLLS.EXE (via the EasyReg main module), a window becomes visible with this information.

The reason for this replacement is twofold:
• The files TNLLS.EXE and TNLLS2.EXE were very big. The web browser version is much smaller in size. (Due to this replacement, the new installation files fit on 13 1.44 MB disks rather than on 15 disks.)
• The new version is easier to upgrade and more flexible. Consequently, it is now more detailed.
• For large horizons, the text box in module VAR.EXE which displays the innovation responses may reach its maximum capacity, causing a fatal "out-of-memory" error. Now you will only get a warning message.
• The chi-square test for cointegrating restrictions on the intercept or trend parameters in Johansen's cointegration test (module COINT.EXE) was incorrect. Effective April 9, 2000, the error has been repaired. Thanks to Dr. Igor Pelipas (Institute for Privatization and Management, Belarus, Minsk) for pointing this out to me.
• Difference between the April 9 and April 20 versions:
• There was a bug in the VAR.EXE module: If you specify and test the same over-identified structural model twice, by clicking the "respecify model" button, the test for correctness of the overidentifying restrictions gave different results. Thanks to Hung-Pin Lai for pointing this out to me. This bug was due to the fact that the first time certain temporary files were overwritten. The side effect of the repair is that the VAR module now allows you to do structural innovation response analysis after having conducted nonstructural innovation response analysis, and vice versa, without the need to re-estimate the VAR.
• I found more bugs in the structural VAR innovation response part of module VAR.EXE. They have been repaired.
• The new template option in the nonlinear regression module (NLLS.EXE) has been disabled: It did not work correctly, and I could not fix it easily. I will try again later this year.
• The module DATA.EXE did not recognize valid data files of format 2 anymore. It does now again.
• If you run Windows at a higher resolution than 640 x 480 pixels, you can now set EasyReg full screen: All the windows, controls, and fonts will be proportionally enlarged, except the menus and picture boxes. This is particularly convenient for laptops, which often do not allow full screen display at a resolution of 640x480 pixels or even 800x600 pixels: if you set these resolutions, on most laptops the actual screen area is reduced in size, leaving a black area around the visible screen. This happens for example with Toshiba laptops. Some other laptops, like the IBM thinkpad, allow for full screen display in 640x480 pixels resolution, but at the expense of blurred and/or ragged fonts.

However, EasyReg is designed for optimal display on a desktop screen area with 640x480 pixels. Therefore, it is recommended to set your desktop resolution to 640x480 pixels, if possible.
• Test runs of EasyReg in full screen mode revealed quite a few cosmetic bugs. In particular, some windows did not line up, and quite a few windows were loaded twice, due to the difference between Visual Basic 3 and 5.
• EasyReg was designed as a single user program: if you activate EasyReg twice you will get a message saying that a previous instance of EasyReg is still running, and the second instance of EasyReg is aborted. Due to a query by Dr. Matteo Manera (Bocconi University, Milan, and Politecnico of Turin, Italy), who wanted to use EasyReg as the main econometric software in his applied econometrics classes, and make EasyReg available to his students via a Windows NT network server, I have redesigned EasyReg such that it now runs in multiple users mode on a network, using a particular shortcut. The EasyReg web page explains the shortcut, and so does EasyReg itself, via the 'Help!' menu. The single user mode is still the default mode, though.
• The new full screen mode and the multiple users mode options have affected all the (38) EasyReg modules, and the redesigned modules are therefore no longer compatible with older versions. Consequently, it is not possible to upgrade previous versions of EasyReg using a small upgrade file, because the upgrade file would become almost as large as the installation file itself. If you want these options, and the bugs fixed as well, you have to re-install EasyReg completely. Of course, future upgrades will be made available again via upgrade files.
• Difference between the April 20 and April 26 versions:
• The previous full screen mode option worked, but I did not like its looks. Therefore, various improvements have been made. In particular, the fonts can now be adjusted in full screen mode. Moreover, the 'Maximize' control box in the upper-right corner of the EasyReg start window has now been enabled. By clicking it, EasyReg will be enlarged to full screen, and (in contrast with most other Windows applications) all the fonts and controls will be enlarged proportionally (which now works better than before), except menu bars and picture boxes.
• In full screen mode, the menu bar may be difficult to read. Therefore, an alternative menu is now available in full screen mode, which can be accessed via a 'Large menus' button. Thus, you no longer need reading glasses to use EasyReg in full screen mode.
• If you run EasyReg in multiple users mode, the title bar will now display the registered name of the user.
• In the unit root module (UNITR.EXE) the same notation p was used for the lag length of the ADF auxiliary regression and the truncation width of the Newey-West in the cases the Phillips-Perron and the KPSS tests. Since my students got confused, the Newey-West truncation width is now denoted by m.
• My previous statement that an upgrade file would be almost as big as the installation file itself was not correct. As requested by an EasyReg user in Brazil (for whom it took 4 hours to download ER32INST.ZIP), I have now added small upgrade files for all the versions since February 1, 2000.
• Difference between the April 26 and May 2 versions:
• If you run EasyReg in multiple users mode, an error number 76: "Path not found" occured if the user's folder specified in the command line contained sub-folders. For example, if EasyReg is installed in network folder N:\EasyReg, and the user's folder is U:\EasyRegUsers\User1, the shortcut

N:\EasyReg\EasyReg.exe U:\EasyRegUsers\User1
would produce this error, while the shortcut
N:\EasyReg\EasyReg.exe U:\EasyRegUsers1
would work fine. The reason was that EasyReg tried to create the user's folder with all the sub-folders at once, which is not allowed in Visual Basic 5. Consequently, the folder U:\EasyRegUsers\User1 will not be made at all, and when EasyReg starts to write files in this non-existing folder, the error involved occured.
This problem has been repaired. The upgraded module EASYREG.EXE is contained in the upgrade file ER32UPG3.ZIP.
• Difference between the May 2 and May 19 versions:
• The modules OLS.EXE, NLLS.EXE, LIMDEP.EXE, TSLS.EXE, and QUANTILE.EXE have been split up in smaller modules, in order to create more memory.
• The module OLS.EXE computed the OLS results twice. This bug has been repaired.
• Difference between the May 19 and August 2 versions:
• The start-up folder can now be created via EasyReg. Previously, only an existing folder could be used as the start-up folder.
• The data table (module DATATAB.EXE) can now also be viewed via Windows NotePad or WordPad, which is faster than viewing the data table via EasyReg.
• The artificial cointegrated VAR data generator in module DATA.EXE has been confined to the bivariate VAR(1) case with cointegrating vector (1,-1)', in order to guarantee that the data generating process y(t) involved is such that (1,-1)y(t) is stationary and (1,1)y(t) is a unit root process.
• The option of including Chebishev time polynomials in GMM models has been removed, because it did not work correctly. Moreover, in selecting the X variables and instruments the chosen dependent (Y) variables are now excluded from the list of potential X variables and instruments.
• A bug in module NLLS.EXE has been repaired. This bug was only effective if the transformation "Multiply" is clicked when you build up the nonlinear model, and in view of the guided tour it appears to be of a recent date.
• Effective July 12, the guided tours have been upgraded. The data files used in these tours are now included.
• The GMM parameter estimates are now re-estimated under linear restrictions if you test for the linear restrictions involved. This upgrade is due to a request by Thomas Osang (Southern Methodist University).
• A new module (BINOMIAL.EXE) has been added to the Tools menu, which computes binomial numbers, and the probabilities of the binomial, hypergemometric, negative binomial, and the Poisson distribution.
• The Help windows in module DATA.EXE which give information about data formats and panel data did not work properly. Similar (non-essential) bugs have been found in a few other modules as well. These bugs have been repaired.
• A bug in the Tobit module has been repaired.
• All the modules have been recompiled in order to check for other bugs. A few were found and have been repaired.
• Differences between the August 2 and August 24 versions:
• A typo in the build-in table of critical values of the ADF test 2 has been corrected.
• The nonlinear least squares module did not always give correct results, due to the fact that not all the computations were done in double precision, in particular numerical differentiation. Now all the computations are done in double precision.
• A non-essential bug in the ARCH testing module has been repaired, and the options to test either ARCH(p) for a single p, or ARCH(m) for m =1,..,p, have been added. Moreover, ARCH testing has now been confined to OLS models only.
• The guided tour for first time EasyReg users on how to import data refers to a file D:\NELPLOS.DAT. It is now explained that the link to file NELPLOS.DAT is actually a link to file DATA.TXT in the EasyReg folder, and that in order to mimic the guided tour you have to copy the contents of file DATA.TXT to D:\NELPLOS.DAT (or C:\NELPLOS.DAT if you do not have a D: drive) via Windows Notepad or Wordpad.
• The 'About' -> 'EasyReg' menu item now links to file EASYREG.HTM in the EasyReg folder, via the default web browser. This HTM file contains about the same information as the EasyReg web page with the same name, but without the grafics.
• Differences between the August 24 and September 28 versions:
• There was a non-essential bug in module ARIMA.EXE: The one-step ahead out of sample forecasts procedure did not stop after the first missing observations at the end of the input file. It does now.
• The module for choosing the fonts has been redesigned. It is now much faster.
• A non-essential bug in the ARIMA data generator in module DATA.EXE has been repaired (in the print-out of the generated model the level of differencing was not indicated).
• Differences between the September 28 and November 28 versions:
• Two errors in the GMM module have been repaired. The degrees of freedom of the Wald test for the validity of the instruments was incorrect: The degrees of freedom was the total number of instruments rather than the total number of instruments minus the number of parameters. See also the upgraded lecture notes on the method of moments. The other error was that the dependent variables could not be selected as X-variables, which excluded dynamic panel data models, and simultaneous equations.
• The Tobit module has been augmented with an explanation of what Tobit models are about, because I got too many e-mail queries about Tobit estimation.
• The transformation module TRANS.EXE has been augmented with the following transformations:
• Dummy = I(x=a)
• 1 if x > a, missing value if not
• 1 if x < a, missing value if not
• 1 if x = a, missing value if not
These transformations have been suggested by Dr. Jose Antonio Ortega Osona. The latter three transformations only apply to cross-section data. They can be used to select the data for a regression model according to the value 'x' of a variable.
• The default value of the parameter 'mu' in the Bierens HOAC unit root test was 5, whereas the value used in the paper was 2. The latter value has been re-adopted as the default value.
• A new module, EASYREGN.EXE, has been added. If you run EASYREGN.EXE instead of EASYREG.EXE, EasyReg will run in multiple users mode. See the updated instructions on the EasyReg web page.
• Upgrades not included in the November 28 version:
• The file DATA.TXT mentioned in the guided tour for first time EasyReg users was not included in the setup package of the November 28 version.
• The ARCH test was incorrect if the model contains lagged variables and/or a subset of observations is used.
• The missing file DATA.TXT and the upgraded module ARCHT.EXE are included in upgrade file ER32UPGR.EXE.

• 
  Back to the EasyReg page
  
  
• 
  Current version of EasyReg