The framework is that of regression with uncertainty about which predictors are in the regression model, with the "model uncertainty" problem addressed in a Bayesian framework. The analysis explores the space of potentially very many models defined by subsets of predictor variables, using the approach detailed in Hans, Dobra and West in Shotgun stochastic search for regression with many candidate predictors (Journal of the American Statistical Association 102: 507-516)

The current code includes both serial and parallel versions, implementing SSS for linear, binary (logistic) and survival (Weibull) regression model frameworks. Examples in the above references illustrate the approach, and the code is accompanied by one example of each of these regressions. SSS is written in C++ and the executables are available for download and use. Inputs and outputs are in simple text files. Full details and all downloads are available under the Serial and Parallel version links above.

Related links and acknowledgements

The parallel SSS software development grew out of earlier code for stochastic search in graphical models. The graphical model SSS methods are implemented in Adrian Dobra's HdBCS code. The current serial and parallel versions of SSS for regression have been substantially developed and we are grateful to a number of Duke colleagues and collaborators for testing, comments and feedback; among others, we particularly thank Jun Zhai for her input and feedback.

The research and development underlying SSS was supported, in part, by the National Science Foundation (grants DMS-0102227 and 0342172) and the National Institutes of Health (grants HL-73042 and CA-112952). Any opinions, findings and conclusions or recommendations expressed in this work are those of the authors and do not necessarily reflect the views of the NSF or NIH.

This software is made freely available to any interested user. The authors can provide no support nor assistance with implementations beyond the details and examples here, nor extensions of the code for other purposes. The download has been tested to confirm all details are operational as described here.

It is understood by the user that neither the authors nor Duke University bear any responsibility nor assume any liability for any end-use of this software. It is expected that appropriate credit/acknowledgment be given should the software be included as an element in other software development or in publications.