H O M E / I N F O / P R O D U C T S / O R D E R / S U P P O R T

Contents / Getting started / FAQ / Recent changes

TSP Newsletter – February 2008

Yes, TSP International still exists and we have been developing and shipping programs.

But we did experience a substantial order backlog around the time of the release of Version 5.0 and many of our loyal customers may be wondering what happened to us. We are very sorry about the delays in response and are determined to do better in the future. NOTE THAT THE BACKLOG WAS CLEARED LAST FALL - IF YOU ARE MISSING AN ORDER, PLEASE CONTACT US.

In order to let you know what we have been up to, we are relaunching our email newsletter with this issue.

News Items:

1. We are now compiling a subscriber list of interested current, former, and potential future users, to be added to the list, go here.

2. TSP Version 5.0 was released last year, see the list of new features and the most up-to-date release notes here.

If you haven't looked at TSP recently, here are some things you may not know about:

Windows, MAC and unix versions have high resolution graphic plots available.

Windows, MAC and unix versions have the entire Reference Manual available as a hypertext HELP system, so you can look up any command easily.

On Windows, TSP runs in the Givewin shell, and is part of the Oxmetrics family of programs, which provides a number of useful features. Jurgen Doornik’s website has information on Givewin and Oxmetrics. We expect to release an Oxmetrics 5 version of TSP shortly.

Here is a sampling of new features in TSP 5.0:

Panel data estimation for linear regression, probit, instrumental variable, and multiple linear regression models.

Robust estimation: censored quantile regression, Kernel estimation

Three-dimensional surface plots in TSP/Givewin

.........and many other changes described here.

3. We are considering development of a number of new features and welcome your input - here are some ideas, but please send your own wishlist to Clint:

Likely:

- Grouped HS-consistent standard errors in all procedures (not just OLSQ, INST,

- Add the (population) weighting option to the commands: LAD, Probit, Logit

- Augmented kernel command allowing for alternative kernel functions (rectangular, epanechnikov, local linear kernel regression with or without a ridge parameter. See Froelich, M. (2004), Finite Sample Properties of Propensity-Score Matching and Weighting Estimators, Review of Economics and Statistics 86(1):77-90.

- Automated LIML/FIML for dynamic panel data, as an alternative to GMM

- FIML on unbalanced data (mostly for panel data) - involves a separate likelihood for each subsample.

- Conversion of stacked panel data to cross section or time series by appending date or ID suffixes and vice versa (wide vs long)

- New functions MIN(a,b)and MAX(a,b)

- MERGE command

- SORT(METHOD=QUICK/BUBBLE) to preserve the order of time series data when doing multiple sorts on different keys (the method affects the resulting order)

- increase max length of filename from 128 to 256 bytes

- read new Excel/Office 2007 formats (zipped XML) - XLSX, XLSB

Unsure:

- CNORM3 option in CDF command

- New functions: beta, igamma

- Nonlinear GARCH - user supplies FRMLs for residual and variance eqns

- Multivariate GARCH.

- Nonlinear quantile regression.

- Multivariate quantile regression. Not sure if this is even well-defined, but has been user-requested.

- Matrix FRMLs (MFRML)

A. matrix algebra (multiply, inverse, quadratic form, det, etc.) on them yields another MFRML or scalar FRML. Then usable in ML.

B. for use in KALMAN for estimating hyperparameters in state space models

C. structural linear models (MOMENTS/LISREL)

- Multivariate ordered probit

- Ordered probit with random effects

- Constraints for nonlinear estimation

- Randomized starting values for nonlinear estimation (to check for multiple local optima).

If you have any questions or comments about TSP please send an email to info@tspintl.com.

Comments or questions about this website should also be sent to the info@tspintl.com. Lost? Please consult the site map.