Regression trees have been used for a range of
applications in different research areas for more than 30 years.
In 1982, Prof. Antonio
Ciampi introduced RECPAM, a flexible and general approach that in the
years to come would have been showing high accuracy and interpretability
in situations frequently occurring in biostatistics.
In 1994, A/Prof Fabrizio
Carinci started the construction of a specialised package to support
the exploration of growing health data-warehouses with RECPAM as a data
mining engine.
An initial version was developed as part of his
research fellowship at Consorzio Mario Negri Sud, 1994-1997. The product
was later refined into RECPAM/SAS, after intense work done at Harvard School
of Public Health and McGill University, 1997-1998.
In the last years, RECPAM/SAS has been successfully
used for the analysis of epidemiological studies and very large health
databases. Its application follows the general rules of decision
trees, which have been used in medicine to reflect diagnostic and therapeutic
strategies and emulate the human cognitive process.
Further to this, RECPAM/SAS produces a prediction
tree, a particular kind of decision tree. The "decision" is to make a specific
prediction given certain clinical characteristics.
This prediction is arrived at by a structured
sequence of yes/no questions concerning the clinical characteristics.
RECPAM/SAS allows to integrate generalised linear
models with survival analysis, merging the logic of clustering with the
probabilistic vision of the dependence of a pre-specified outcome from
a set of potential predictors. Details on the software, examples of its
application and the software itself may be examined following the links
on the left.
Although the development of RECPAM/SAS as such
stopped in early 1999, its foundations have been reused at Monash University where additional source
code has been put together for an improved parallel, web-enabled version, under the banner of the
REIGN
project.
