<Next version: PcGive 12>
<Previous version: PcGive 10>
PcGive 11: enhancements and solved problems
Problems in PcGive 11.0 that have been fixed in PcGive 11.1
- Switched off automatic instability tests for cross-section
- GIVE estimation: fixed small mistake in constant of
log-likelihood (reported in progress).
- Panel and Arfima batch would not used the specified
estimation method but default to the first one.
- Storing Principal Components in database goes wrong when
estimation is over a subsample. The dialog is also hidden behind the model
- Automatic output of cointegration tests in system estimation
leads to infinite recursion in Output, resulting in a stack overflow.
Problems in PcGive 11.0 that have been fixed in PcGive 11.04
- OxMetrics help index removed all entries with duplicate
links. The drawback was that many useful keywords disappeared.
Batch entries where missing from the PcGive help index.
Arfima and Garch chapters were wrong in help index.
- Options/Additional output: not printed for Cointegrated VAR
- F1 Help for GARCH/Arfima forecast dialog doesn't work.
- Arfima/GARCH/Logit/Panel didn't use correct graphics
window names. Should be Model for Graphic Analysis, Forecasts
for Forecasts, with the others (such as recursive) going to Other.
- Single-equation, Cross-section, Multiple-equation batch
commands testgenres, testlinres, testres
gave a run-time error.
- I(2) tests on model with restricted constant and
additional variables restricted to the cointegrating space generates
New in PcGive 11
- Now a module in the new OxMetrics 4 system.
- Improved integration with OxMetrics front-end: PcGive no longer
appears as a separate program, but is accessed from OxMetrics.
- Dialog design has been improved.
- Addition of Monte Carlo analysis of linear dynamic econometric
models. Simple AR(1) and regression experiments can be designed
and run interactively, displaying graphs as the Monte Carlo progresses.
More advanced experiments can be created interactively, and then run
using Ox Professional.
Problems in PcGive 10.4 that have been fixed in PcGive 11.0
- Unrestricted system/cointegrating test: expected
lag 1 variables in the same order as the dependent variables.
(Would give wrong answer otherwise.)
- Components graph in Cointegrated VAR: labels of variables
restricted to cointegrating space are wrong.
- Panel estimation:
GMM instruments are kept in the model when switching from
dynamic to static estimator. The GMM instruments are listed in the output,
but the results are different when going to static directly (the GMM
instrument dialog is only offered in dynamic estimation, so they can only
be deleted there).
This file last changed .