Microfit is supplied with an interactive econometric text which gives details of the econometric methods that underlie the options in the package. It also contains numerous tutorial lessons and exercises in a large number of areas covering univariate and multivariate forecasting of financial series such as stock prices, interest rates and exchange rates.
Microfit is an ideal tool for macro-econometric modelling, incorporating most up-to-date econometric methods.
Microfit Navigation Bar

The GARCH options in Microfit allow investment banks and portfolio management, companies to model and forecast volatility of share prices, interest rates and exchange rates in a matter of minutes.
Microfit 4.0 represents a major advance over the earlier versions of the package. It makes more intensive use of screen editors and window facilities for data entry, model specification, and easy storage and retrieval of data and results files.
Microfit 4.0 accepts ASCII and binary data files, and Excel 4.0 Worksheets, and other data files with a variety of formats such as comma delimited (CSV), PRN (Lotus print files), TXT and AREMOS (TSD) files. It readily allows for extension, revision, and merging of data files. Data on Microfit's workspace can be exported to spreadsheet packages in the CSV format, and to the AREMOS package in the TSD format.
With this new version you can run regressions up to 100 regressors and 3,000 observations. You can also move readily between drives, directories, and subdirectories for retrieving and saving data input and output files. Scrolling within a results screen is also possible. All files created using Microfit 3.0 can be used in Microfit 4.0.
Microfit 4.0 now comes with an extensive system of help facilities, providing easy on-line access to the Microfit manual. The graphic features of the package have been upgraded further, and now allow the graphs produced in Microfit to be readily imported into word-processing packages or printed on any printer supported by Windows.
New Single-Equation Options:
Maximum likelihood estimation of regression models under a variety of conditionally heteroscedastic error specifications, such as ARCH, GARCH, GARCH in mean, absolute value GARCH, absolute value GARCH in mean, exponential GARCH, exponential GARCH in mean. The ARCH and GARCH models can be estimated for two different specifications of the conditional distribution of the errors, namely normal and the Student-t distributions.
LOGIT and PROBIT estimation.
Phillips-Hansen's Fully Modified estimation
of cointegrating relations.
A new Autoregressive-Distributed Lag (ARDL) approach to estimation of cointegrating relations. This procedure allows you to choose the order of the ARDL model by means of the model selection procedures such as Akaike, Schwarz, and the adjusted R-square statistic. This approach also allows for inclusion of time trends, seasonal dummies and other deterministic/exogenous regressors in the cointegrating relation.
Non-nested tests of linear versus log-linear models, and level-differenced versus log- differenced models, and other non-linear specifications of the dependent variable.

©Oxford University Press Ltd & Camfit Data Limited, 1996.

For data analysis, Microfit 4.0 has a large number of additional time series and econometric features. These include:
New Functions and Commands
New Functions Included in Microfit Are:
RATE(X) function, which computes the
percentage change in variable X
MEAN(X) function, which computes the mean of X
STD(X) function which computes the
standard deviation of X
MAV(X,p) function which computes a pth
order moving average
Hodrick and Prescott Filter
New Commands in Microfit Are:

REORDER X which activates a complete
reordering of the observations on the workspace according to the ordering of X. This command is particularly useful for use of Microfit in cross-sectional analysis
RESTORE, which restores the ordering of the
observations to their original state before the use of the command REORDER

New System Equation Options:
Unrestricted VAR estimation.

Automatic order selection in VAR using Akaike, Schwarz, and likelihood-ratio procedures.
Granger non-causality tests in the VAR.
Orthogonalized (a la Sims) and Generalized Impulse response analysis in VAR models. The generalized impulse responses are new and, unlike the orthogonalized responses, do not depend on the ordering of the variables in unrestricted VAR models.
Orthogonalized and Generalized Forecast Error Variance Decomposition in unrestricted and cointegrating VAR models.
Estimation and hypothesis testing in systems of equations by the Seemingly Unrelated Regression Estimation (SURE) method and restricted SURE method.
ML estimation and hypothesis testing in systems of equations subject to parametric restrictions. The restrictions could be homogenous or non-homogenous, and could involve coefficients from different relations (i.e. cross-equation restrictions).
Long-run structural analysis based on a cointegrating VAR. This estimation procedure allows you to estimate and test more than one cointegrating relations subject to identifying and over-identifying restrictions on the long-run (or cointegrating) relations. The restrictions could be homogenous or non-homogenous, and could involve coefficients from different cointegrating relations. It also allows analysis of sub-systems where one or more of the I(1) variables are exogenously determined.
Generalized and orthogonalized impulse response analysis and forecast error variance decomposition in cointegrating VAR models.
Computation of persistence profiles for the effect of system-wide shocks on the cointegrating relations.
Computation of multivariate dynamic forecasts using unrestricted or restricted VAR models.