Simulation

 

The directory where simulation output is written. When you run a simulation SVAR automatically saves the results in 2 files; a summary in text format and the full simulation in a MatLab binary mat file. This option allows you to select exactly where such files will be stored. By default they will be saved in the subdirectory "simulate" to the directory where SVAR is located.

 

Presents all, if any, Matlab MAT Files in the simulation output directory. A dialog allows the user to select the files that should be deleted.

 

Presents all, if any, Text Files in the simulation output directory. A dialog allows the user to select the files that should be deleted.

 

Presents all, if any, Matlab M Files in the simulation output directory. A dialog allows the user to select the files that should be deleted.

 

SVAR will use parametric bootstrapping by default. This means that standard normal errors are drawn when creating pseudo-data, where the estimated covariance matrix is used to transform the standard normals to pseudo-data errors. If the checkbox is not checked, then SVAR switches to using non-parametric bootstrapping, i.e., to re-sample among the estimated residuals when creating pseudo-data. The selected value can always be over-ridden on the individual bootstrap dialogs.

 

SVAR can perform block bootstraps under non-parametric bootstrapping, i.e., when drawing from the estimated residuals. To activate block bootstrapping you must select a value for the block size in excess of 1. The maximum number for the block size is 12.

 

By default SVAR does not save and thus allow for reuse of bootstrapped data. By check marking this feature the default is changed to such behavior. Whatever value you select you can temporarily change the state of this option on each individual bootstrap dialog.

 

The default behavior for SVAR when generating pseudo-data through the parametric bootstrap routine is to use a Choleski decomposition of the covariance matrix Ω when creating new residuals though its random number generator. This option makes it possible to apply a symmetric decomposition instead, via the eigenvalues and eigenvectors of Ω.

 

Select how the state is determined when drawing random numbers. The default is MatLab's own default. Options including clock sum are based on summing the 6 elements in the MatLab generated vector clock (year, month, day, hour, minute,seconds).

 

The number of times to repeat a generation of data or computation of a test statistic (default is 10,000 times or 9,999 for bootstraps).

 

The number of realizations of a random variable for each replication (default is 500 realizations).

 

When SVAR graphs simulated distributions it either uses MatLab's interp1 function or kernel density estimators. The interp1 routine has 4 options (nearest neighborhood, linear, cubic spline, and cubic). The kernel density estimators have 8 different kernels: normal, Epanechnikov, rectangular, triangular, bi-weight, tri-weight, Laplace, and logistic. For simulations with relatively few replications the kernel density estimators usually provide better approximations than the other interpolation functions. By default the Kernel density Epanechnikov function is used by SVAR.