GARCH Tool

User interface used to fit and evaluate generic GARCH models (AR, MA, ARMA, GARCH) to loaded data.

The data may be loaded from,
- an Excel file that contains a time series, with dates down the first column and data down the second column. The first row of the spreadsheet is assumed to be column headers.
- a MATLAB Workspace variable. The variable must be a structure with a field called dates that is a vector of date numbers such as those generated by the datenum function, and a field called data which must be a numeric vector the same length as the dates. Any other fields in the structure are ignored.

Once data is loaded the UI allows the user to,
- process the raw input data by selecting (sub-)ranges, convert to returns, and differencing the data up to twice.
- view the Auto Correlation and Partial Auto Correlation of the selected data.
- View the Box-Jenkins Stationarity test for 0-12 lags
- Specify and fit an ARCH/GARCH model.

Once a model is specified and fit to the data the UI allows the user to,
- view the fitted model parameters
- view the Ljung-Box Q test on the standardized innovations and the standardized innovations squared
- view the autocorrelation of the standardized innovations and the standardized innovations squared
- plot predictions out into the future
- plot simulations out into the future

At any stage the raw data, processed data, the model, the prediction and the simulation data may be exported to the MATLAB Workspace. (If a model has not yet been fitted and/or a forecast not generated then those fields will be empty in the exported data set.)

There is limited documentation and those unfamiliar with GARCH modeling should refer to some of the examples within the Econometrics Toolbox documentation.