Mittwoch, 7. September 2011

Parametric estimation for GARCH-class models with MS Excel and nonlinear GRG-Solver

There are quite a lot of software packages with the ability of estimating GARCH-class models, but some of them are too expansive (e.g. EViews) and the most are very difficult to handle. Therefore I create an open source Excel-Tool which is prepared for ARCH(1), GARCH(1,1), IGARCH(1,1), GJR-GARCH(1,1) and EGARCH(1,1). Other models will be added later. All parametric estimations are using the likelihood function in combination with a nonlinear GRG algorithm (Solver-Add-In for MS Excel is needed).

Everything you have to do is:
  1. Download and open the Excel spreadsheet 
  2. Import or copy some time series into the sheet “time series” with latest rate above (some are already available)
  3. Go to the sheet “Estimate”
  4. Chose model and time series and click the start button
The solver is now trying to maximize the likelihood function (MaxL), and gives back the corresponding parameters. The field statistics gives insights regarding the model quality. The second test statistic in the field explains the degree of clustering squared returns before the modeling. The fourth explains the degree by use of the model. It should be lower than the second test statistic, and ideally be close to 0.5. At this point, absolute randomness is reached. The first and third test statistic explains the clustering of the direction of price changes. Both should be close to 0.5, since these are assumed to be random. Have fun and give feedback!