robust


The software in this page provides minimum and maximum values of the posterior mean of a function recorded in a posterior simulator file, as the prior is changed from its original specification. The prior changes in such a way that it has a kernel that is bounded above by a multiple k of the original prior density, and bounded below by a multiple k-inverse of the original prior density; k exceeds 1.

More details are provided in the files robust.ps and robust.pdf in this page. The text in these files also provides further references to the literature.

Input for robust consists of three records. The first record consists of three integers: the number of alternative values of k to be used; the number of burn-in iterations to be discarded in the computations; and the element of the simulation vector corresponding to the function of interest. (If this function is not included in the simulation vector it can be created using the software reweight.) The second input record contains the alternative values of log(k) used; log(k) must be between 0 and 8. The third input record is the name of the posterior simulation file.

Computation time for this algorithm is very fast. It depends only on the number of iterations of the posterior simulator, but it only takes a few seconds.

A description of the methodology can be found in the BACC manual.


Languages




C routines

To run the program:


Files needed to create executable:

Library Files needed to obtain executables:



Fortran Routines

To run the program:

Files needed to obtain the executable:

Libraries needed: This routine uses IMSL. The user needs to link to IMSL or to the libraries described in LIBRARIES to generate executable.


Gauss

To run the program:

Files needed:


Mathematica

To run the program:

Files needed to obtain the executable:



Matlab

To run the program:

Files needed to obtain the executable:



SPlus

To run the program:

Files needed to obtain the executable:


Samples