LETTERS TO THE EDITORS
Statistically Significant
To the Editors:
While no scientist should disagree with Hugh G. Gauch, Jr.
("Winning the Accuracy Game," March-April) regarding the
need for sophisticated statistical approaches to scientific
problems, many times there is an education gap that prevents such
techniques from being used to full advantage.
For many scientists, statistics is not a discipline that is high on
their horizon, and no requirements for such courses exist in their
curriculum. This leaves their future ability to use such tools to
self-study, or perhaps an elementary course that may tend to give
scientists a limited ability to design their programs or analyze
their results.
I have always been astounded by the use of statistics by some
political and social scientists, who are unable to properly
interpret the results but nonetheless have their findings published
in the popular press, to the discomfiture of their more
skilled-in-statistics colleagues. Statistics is a necessary tool,
but in the hands of the ill-trained, it is a dangerous one.
Nelson Marans
Silver Spring, MD
Dr. Gauch responds:
I certainly agree that many scientists receive an inadequate
education in statistics. I suspect that there are several problems.
One is that science is rapidly changing, particularly in the sheer
quantity of data that many scientists now accumulate.
Fortunately, where I am at Cornell, the chair of Statistical
Sciences has been involved with administrators to gather information
on what kinds of data professors and graduate students are
collecting and what kinds of data analysis and hypothesis tests they
need, so that statistics courses can be updated and tailored to fit
the students' actual needs. Among the results of this effort are
more emphasis on multivariate analysis and the Bayesian paradigm.
Another problem, which my article already mentioned, is that
statistical procedures or advances in one scientific specialty may
be imported very slowly into other specialties with similar data
structures and problems. In this regard, general science
publications such as American Scientist have a vital role
to play, because their readership spans countless disciplines and specialties.
But one bright hope is that a striking success story—like a
remarkable gain in efficiency and accuracy—can attract attention.