EasyReg International, 2005 versions
Differences between the versions
November 17, 2004 and August 5, 2005:
- Additions and improvements
- A link to the new mirror site in Poland has been added, under menu WWW.
- In the summary statistics module the option has been added to compute these
statistics only for the observations for which none of the selected variables
have missing values. This option only applies of course if you select more than
- A new module (CVSSUBSAMPLE) has been added, under menu File > Choose an input file.
This module allows you to take a subsample from a CSV file that is too large
(more than 32767 observations)
for EasyReg to handle directly. Moreover, the existing modules DATACSV and DATAOLD
will no longer crash is you try to import a data file with more than 32767
observations, but instead they will now refer to the new module.
- The "What to do next?" module will now ask for confirmation when you click "Done".
The reason is that I forgot a few times to conduct particular options after hours
of computations, so that I had to start all over again.
- The module SURVIVAL [PDF] which estimates
the interval-censored proportional hazard model has now four
more options for the baseline hazard function. Moreover, the output now displays
the Akaike, Hannan-Quinn and Schwarz information criteria.
Furthermore, this module now comes with a guided tour.
- The module for upgrading EasyReg (via the WWW menu) now gives the advice
to upgrade via the upgrade files on the EasyReg
download web page if the number of newer files is larger than 15.
- A new module, SNPSURVIVAL [PDF],
has been added. This module estimates an interval-censored mixed
proportional hazard model, where the unobserved heterogeneity distribution is
estimated semi-nonparametrically. This approach is based on my working paper:
This module is available via Menu > Miscellaneous modules, for cross-section data only.
A guided tour is under construction, but will be similar to the guided tour on
- Bierens, H.J. (2005), "Semi-Nonparametric Modeling of Densities on
the Unit Interval, with Applications to Censored Mixed Proportional Hazard Models
and Ordered Probability Models: Identification and Consistency Results" [PDF].
- Two new modules have been added, SNPSURVIVAL1 and SNPSURVIVAL2. Both modules
are available via Menu > Miscellaneous modules, for cross-section data only. Guided
tours are not yet available, but are planned.
- Module SNPSURVIVAL1 [PDF] estimates a
univariate mixed proportional hazard model for a duration T, without bracketing, except
for right censoring. Censoring is indicated by a dummy variable C:
given that T is only observed over the period [0,Tmax],
C = I(T>Tmax),
where I(.) is the indicator function. The observed T is assumed to be equal
to Tmax if C=1. The unobserved heterogeneity is integrated out in the
same way as in Bierens (2005) [See the reference above], namely via a
semi-nonparametric distribution function H(u) on [0,1]. However, you also
have the option to specify a Gamma distribution for the unobserved heterogeneity.
- Module SNPSURVIVAL2 [PDF] estimates a bivariate mixed proportional hazard
model, without bracketing, where the two durations T1 and T2 are only observed
as T = min(T1,T2), together with a discrete variable D which is 1 if T2>T1
and 2 if T2 < T1: D = 1 + I(T2>T1), where I(.) is the indicator function.
Again, right censoring is indicated by a dummy variable C = I(T>Tmax).
The observed T is assumed to be equal to Tmax if C=1. Conditional on a vector
X of covariates and a common unobserved heterogeneity variable V, the durations
T1 and T2 are assumed to be independent. The random variable V will be integrated
out in the same way as in Bierens (2005), via a semi-nonparametric
distribution function H(u) on [0,1]. However, you also have the option to specify
a Gamma distribution for the unobserved heterogeneity variable V.
This module is intended for the revision of:
- Carvalho, J. R., and H. J. Bierens (2002), "A Competing Risk Analysis of
- The "What to do next" module (NEXTMENU) has been upgraded in order to
accommodate these new modules.
- A new module, VARFORECAST, has been added, which conducts
one-step-ahead and recursive out-of-sample forecasting on the basis
of a VAR model. This module is available via Menu > Multiple equation models.
- A new module, ARIMAMODSEL, has been added (under Menu > Single equation models)
which automatically selects an ARIMA model based on the Akaike, Hannan-Quinn and
Schwarz information criteria.
- The ARIMA module and the modules that extend an OLS regression to models with
ARMA and/or GARCH errors (NEXTARMA and NEXTGARCH) are now endowed with the default option to
automatically restart the simplex iteration until the objective function does not
change anymore. Moreover, the interfaces of these modules have been improved.
- The NEXTGARCH module now has the option to plot the in-sample GARCH error variances,
and it automatically generates out-of-sample forecasts of the GARCH variances.
- After re-estimating a model with GARCH errors, you now have the option (in the
"What to do next?" module NEXTMENU) to write the in-sample GARCH variances to the input file.
- A new module, SURVIVAL1 [PDF], has been added (under Menu > Single equation models). This
module estimates a right-censored proportional hazard model without unobserved
- This module and the other survival analysis modules now allow you to change the start values of the
parameters of the systematic hazard. Moreover, in all these modules you now have the
option to run them in batch mode, in order to accommodate lage jobs that need to run
- Numerous cosmetic upgrades.
- Bug repaires
- Due to a bug, some critical values of the Breitung cointegration test
were set equal to zero. Thanks to Dr. Brian Lucey for pointing this problem
out to me.
- The upgrading of EasyReg and it guided tours via EasyReg did not work under Windows XP.
It appears that the modules involved only work under Windows XP if they are compiled
under Windows 98.
- Some non-essential (cosmetic) bugs in the user-defined nonlinear models modules (ML, NLLS, NLGMM)
have been repaired.