TU13
Bulletproofing and Knowledge Encapsulation in Statistical Macros
    Invited
John K. Troxell
Merck & Co., Inc.
Abstract:
Most SASŪ programmers agree that SAS macros are a wonderful
tool for automating repetitive tasks. More controversial is
bulletproofing (making a macro robust to incorrect parameters or
data). Less commonly considered is the potential that macros
offer for
delivery of statistical expertise. Statistical knowledge can
be built into macros with advanced rules for validating macro
parameter values, implementation of good statistical practice,
implementation of otherwise unavailable statistical methods, or
delive
ry of data-driven advice and interpretation in English. Pros
and cons of bulletproofing and knowledge encapsulation are
discussed, and some specific techniques are presented. Many of
the techniques are useful in non-statistical macros as well.
Examples are
drawn from macros for stepwise regression analysis
and analysis of covariance.
Biography:
John Troxell is a Senior Statistical Programming Analyst at Merck Research Laboratories. Prior to joining Merck, John was a SAS programmer and statistician at Exxon Research and Engineering Co., and then an independent SAS consultant. He has been programmi
ng in SAS since 1978, and still has a copy of the SAS 76 manual. John holds MS degrees in Horticulture and Statistics from Rutgers and a BA in English from Yale.