The main aim of this project was to demonstrate the effectiveness of an Open Grid Services Architecture (OGSA) component-based approach to middleware for handling one of the large-scale statistical modelling problems currently confronting e-social scientists. We extended the software package SABRE to the context of multiprocess multilevel analysis and provided a parallel version of this extended version as an R object. We also explored the possibility of a Web portal to SABRE and SABRE within R.

SABRE is a software package, written in Fortran, which was designed to model recurrent responses for a collection of individuals or cases. Standard generalised linear models can be fitted as well as various mixture models with random effects. We compared the performance of the GRID-enabled version of SABRE with that of the current brand leaders for this type of data, i.e. Stata and gllamm. We illustrated our approach on 16 data sets (of different sizes and complexity). The largest difference occurred in an example with a 2 Gb data set. In this example Sabre (with 16 processors) was over 2000 times faster than Stata 9 and at least 8000 times faster than gllamm.

Start date
30 September 2003
End date
29 September 2005
Grant holder
Professor Robert Crouchley
Professor Roger Penn
Dr Robert Allan
Professor Gordon Blair
Professor Geoff Coulson
Professor Steve Bradley
Dr Damon Mark Berridge
Grant amount
Grant reference
Social Stats., Comp. & Methods
Statistics, Computing and Methodology
Grant type