TSP Features
A complete list of TSP commands is also
available, as well as a list of features added to
TSP 5.0.
- Convenient loading of raw data from free format, fixed format,
binary, Lotus or Excel files
- Flexible transformations of series using algebraic
expressions, with a full set of operators and functions
(transcendental, probability, rounding, etc.)
- Display and save the original and transformed time series;
print a table; graphics on DOS, Windows
and MacOS; write to external files in many formats
- Single equation estimation using a variety of techniques:
- OLSQ Ordinary least squares (with extensive diagnostics)
- 2SLS Two stage least squares (instrumental variables)
- LIML Limited Information Maximum Likelihood
- AR1 Regression with first-order serial correlation
- PDL (polynomial distributed lag) variables can be included in all
these methods, and Shiller distributed lags can be included in
OLSQ
- LAD Least Absolute Deviations regression - minimizes sum of
absolute values of residuals, instead of squares of residuals.
A robust estimator. Handles quantile and censored quantile estimation.
- LMS Least Median of Squares regression - minimizes the median
squared (or absolute value) residual. Another robust estimator,
useful for outlier detection.
- Retrieval of coefficient, residuals, predicted dependent
variables, and other test statistics from these estimators for
later use
- Estimation of equation systems by simultaneous methods:
- LSQ least squares, minimum distance, nonlinear least
squares, multivariate regression and three-stage least squares,
both linear and nonlinear
- GMM Generalized method of moments estimation (multiple
equation, nonlinear, panel data)
- FIML Estimation of a complete nonlinear simultaneous model
by the full information maximum likelihood method
- Extensive hypothesis testing facilities available following
any estimation command.
- Simulation, either dynamic or static, of an estimated
nonlinear or linear simultaneous model
- Time Series methods:
- BJ Box Jenkins identification, estimation and forecasting
- KALMAN Kalman Filter (State Space) estimation
- ARCH ARCH, GARCH, GARCH-M models of autoregressive
conditional heteroskedasticity
- VAR Vector autoregressive models
- COINT for unit roots, Engle-Granger and Johansen trace tests for
cointegration
- Maximum likelihood and qualitative dependent variable
estimation:
- PROBIT Binary probit model
- TOBIT Tobit (0/+) model
- SAMPSEL Sample selection model (generalized Tobit)
- LOGIT Conditional, multinomial, and mixed logit for two or
more choices
- ORDPROB Ordered probit model
- INTERVAL Interval regression (ordered probit with
category boundary values known)
- POISSON Poisson model for count data
- NEGBIN Negative Binomial models for count data
- ML User-programmed likelihood functions with automatically
generated analytic derivatives, examples for Box-Cox, frontier
production, nested logit, switching regression, bivariate probit
- Panel Data estimation, fixed and random effects for balanced
or unbalanced data
- In addition, TSP's special features give user flexibility in
extending the program:
- Ease in handling of multi-sectoral data and equations
- Full matrix algebra
- Control of program flow by logical expressions and tests,
GOTO statements, and loops
- User-written procedures, callable with arguments
- Random number generation for many distributions: normal,
multivariate normal, Cauchy, Student's t, Laplace, exponential,
Poisson, negative binomial, gamma, uniform, empirical (bootstrap)
- Extensive library of TSP procedures, developed to solve
special econometric problems.
If you have any questions or comments about TSP please send an
email to info@tspintl.com. Comments or questions about this website
should be sent to the webmaster. Lost? Please consult the site map.