New in Version 4.26

• Panel data analysis, including OLS and IV for fixed effects models, and
  GLS/ML for random effects models.

New in Version 4.25

• Bootstrap test of I(0).
   
• "RGB" colour-coded scatter plot option - symbol colours depend on
  the observations' position in the sample.
   
• Many small coding improvements and fixes.

New in Version 4.24

• Improved facilities for running and sharing Ox code. Code files now
  bundled with exported settings, and viewable from the Help menu.
   
•  EDF files for tabulations of different sample sized can now be merged.
  p-values are generated by interpolating the tabulations according to
  actual sample size.

NOTE: File formats for these features are changed! Files created with
previous versions of TSM must be modified.

New in Version 4.23

      HAC plug-in bandwidths and prewhitening. 
      New plotting features, including forecast fan plots, choice of line widths,
      line style selection feature with 8 selectable combinations of colour/pattern
      width and symbol.  

New in Version 4.22

Kiefer-Vogelsang-Bunzel (KVB) inconsistent variance matrix estimation.

New in Version 4.21

• Version 4.21.01-01-07 offers new nonlinear time series features

        1) Nonlinear moving average (SPS) model, implementing the STOPBREAK
            and STIMA models as special cases.

         2) Coded formulae can now be recursive, including the lagged value as an argument.

         3) Extended options for coded equations, including codeable nonlinear ECM
             and nonlinear MA functions.
 
         Also several fixes, and code and documentation improvements. The
         programming manual has now been fully revised to incorporate new modelling
         features.  

• Version 4.21.28-02-07 Features a model description field, where users can
  enter details and comments for later reference. Alternatively, a model description
  can be generated automatically in TSM's scripting language.

New in Version 4.20

• Version 4.20.27-09-06 includes three new features:

         1) Nyblom-Hansen model stability test; full model and single parameter
             statistics optionally computed.

          2) Local Whittle ML estimator for the long memory parameter, adding to
              the currently implemented log-periodogram regression procedures.

          3) Automatic regressor selection procedure; computes the models containing
              all subsets of a specified baseline specification, and prints the result that
              optimizes the chosen selection criterion, Schwarz, Hannan-Quinn or Akaike.
              The baseline set can be either all the regressors, or selected Types.
             
               There have also been a number of cosmetic changes to the interface, and
               numerous code improvements and fixes. 

• Version 4.20.07-11-06 implements coded nonlinear forms for equilibrium relations.
   Implementation of supplied EDF option for critical values extended to Ox-coded
  test statistics.
  Text files can now be loaded into the results window for editing.
  "Restart" option, for instant incorporation of changes to user's Ox code - allows
  program to be closed and re-opened (with new code recompiled) by one click.
  Also, several bugfixes.  
   

• Version 4.20.12-11-06 adds a feature to randomize simulation models, using the
  coded equation feature. The main application is to integrate over a range of
  alternative in test power evaluations. There are also several refinements and fixes,
  including a fix of the 'drag-select' feature for model and variable lists.    

New in Version 4.19

• Version 4.19.25-07-06 is a major upgrade with significant new capabilities.
  The most important ones are as follows:
  
  1) A formula parser, which is used to implement three new features: algebraic
  data transformations, nonlinear parameter restrictions (replacing the existing
  procedure, although the restriction format is effectively identical) and nonlinear
  equation estimation. The last feature complements the existing capacity to
  compile Ox code with the program; not quite so flexible, but very much easier
  to implement.
  
  2)  Test statistic tabulations,  in the form of empirical distribution functions with any
  number of bins, can be loaded into the program from spreadsheet files and used to
  compute p-values. These files can be created by the Monte Carlo module, so that
  any test can be tabulated by simulation. Think of this as a way to obtain better, or
  more extensive, tabulations of nonstandard tests than are available from published
  sources.  Also, to tabulate entirely new nonstandard tests, or more generally as an
  alternative implementation of the bootstrap. The tabulations can of course be used
  in Monte Carlo simulations of alternative hypotheses (having been generated from
  simulations of the null hypothesis) to yield estimates of true (size-corrected) test
  powers. 
  
  3) Several upgrades to the user interface including redesigned dialogs, and two
  user-assignable toolbar buttons with a more flexible assignment method. One new
  option, with classroom use in mind, is to simplify the interface by hiding/disabling
  the numerical optimization features.
  
  4) Completing the implementation of the Harris-McCabe-Leybourne (2006) long
  memory test, which was introduced provisionally in 4.18.      
  
  Among more whimsical changes, note that the space shuttle is now taking off, not
  cruising.

• Version 4.19.08-08-06 adds pos( ) and neg( ) functions to the formula parser
  (data transformation and and equations). These can also be used to implement
  max(.,.) and min(.,.). Also, variable bandwidth log-periodogram bias test. 

• Version 4.19.20-08-06 adds a Fourier bootstrap option, and discrete Fourier
  transform command to Data Transformations. Also, changes to the Linear
  Regression and Dynamic Equation dialogs (now separate buttons to open the
  supplementary dialogs). Plus some code fixes and refinements, and revisions to
  the documentation.  

• Version 4.19.04-09-06 implements a Gnudraw upgrade, improving date handling
  and suppressing the annoying mouse pointer coordinates. Also, interface
   improvements with "Options" buttons for quick access to options from the
   Linear Regression, Dynamic Equation and Data Graphic dialogs. Also fixes
  a minor bug in the 20-08-06 release. 

New in Version 4.18

• Version 4.18.01-06-06 runs under Ox 4.0. It features mainly bugfixes and minor
   code improvements, the chief of which is a redesigned Options / Forecasts dialog.
  It appears that on some Windows installations, launching the program by
  double-clicking on a TSM file did not work. This issue has now been
  addressed, but note that it will require a full uninstall/reinstall to
  implement the fix.

• Version 4.18.21-05-06 has two new features.
  1)   Whenever you have a dialog with a variable list open, you can plot your
  selected variables with just two clicks: click on the list, then click the
  Chart button on the toolbar. The Chart button now has a dual function. It
  opens the data plotting dialog if no variable list currently "has the
  focus", and operates the new shortcut otherwise.
  2) TSM can now read Excel-format dates for daily data. If your .xls file has
  a date-formatted column as Column A, these individual dates, in default
  YYYY-MM-DD format, will appear in the Sample dialog and the output. They can
  also be used in a variety of formats to label plot axes.
  The "Stop" button that now appears while calculations are running
  (v4.15-05-06) supplies the functionality previously given by the "Run"
  button. There is still an unresolved timing issue involving the JRE, that
  can cause a freeze-up if the "Continue" option is selected repeatedly, but
  this would seem a rare enough occurrence to tolerate.
  A number of other tweaks and improvements too. The graphing routine should
  now work better with or without daily dates. The File / Settings / Export
  function did not always work quite as intended, and has been redesigned. It
  should load existing exported .tsm files OK, but remake any mission-critical
  files as a precaution.

New in Version 4.17

• User can now interrupt program operations at any point. Progress indicator for numerical optimizations.
• Enhanced code for setting model specifications. 
• Enhanced Monte Carlo module, allows multiple runs.
• 'User-supplied statistic' option.

New in Version 4.16

• Start the program by clicking on a settings (.tsm) file in Windows Internet Explorer.
• Enhanced Export command, new Import Models command.
• Improved installation routine preserves user customizations
• Sieve AR bootstrap option, with regular and wild bootstrap variants.

New in Version 4.15

• Subsampling: an alternative to the bootstrap for resampling inference. See the new Options / Simulation and Resampling dialog.
• Saikkonen-Stock-Watson efficient estimator for cointegrating relations.
• Redesigned interface, with new Dynamic Equation dialog, "floating" sample selection dialog and many small improvements.
• User-assignable toolbar button.

New in Version 4.14

• New settings format means that settings files are fully upgradeable - upgrading is now seamless.
• Error recovery dialog, with re-start button.

New in Version 4.13

• Test or impose general nonlinear parameter restrictions.
• Code any test statistic in Ox, based on estimation outputs.
• Chow tests in linear regression implemented .
• "sup-F" tests enabled (maximum statistic over grid of nuisance parameter values). 
• Enhanced Data Transformation and Editing dialog for easier option selection.
• Redesigned Values dialogs - all equation values in one window.

New in Version 4.12

• Direct Conditional Correlation (Engle 2002) multivariate GARCH specification.

• Automated Wald test for joint significance of all regressors.

New in Version 4.11

• BEKK (Engle and Kroner 1995) multivariate GARCH specification.

• Windows Installer copies files and sets short-cuts automatically.

New in Version 4.10

• Monte Carlo experiment option. Any model supported by the program (including user-supplied code) can be simulated, and the same, or a different, model fitted to the artificial data. Results reported include estimator moments, bias/RMSE, quantiles and p-value EDFs of test statistics. Also, kernel density/histogram plots of replications
• Model Manager - stores and retrieves complete model specifications, including parameter values, forecasts, plots etc. This feature makes it easy to work on two or more projects in the same session, to save and recall a complete record of estimations, as well as facilitating Monte Carlo experiments.  
• Probit and logit models for binary data.
• Poisson and negative binomial models for count data.
• General error distribution (GED) maximum likelihood.
• Nonlinear error correction mechanisms, incorporating ESTAR and asymmetric AR models 
• Extended explained regime-switching option.
• User-supplied likelihood functions. 
• Polynomial distributed lags.
• Durbin-Watson statistic.
• Common factor test.
• Improved graph display with GnuDraw 4. No more annoying 'pop-up' window to shift!
• Statistical tables look-up and density plots.
• Data descriptions feature.   

NOTE: File and folder names have changed, so do not attempt to update an existing installation of TSM40. Install the complete package in a new folder [ox_home]\packages\tsmod4\.
.

New in Version 4.06

• Log-periodogram regression option with Geweke-Porter-Hudak and Moulines-Soulier methods.
• File menu options for saving individual listings and graphics to disk.
• New format for estimation results, now including parameter t ratios.