I almost didn’t see it. On the library bookshelf, wedged
between heftier hardback tomes, was a little paperback booklet titled How to Test Normality and Other
Distributional Assumptions, by Samuel S. Shapiro. It was so thin that the
title didn’t fit on the book’s spine.
Nevertheless, I picked it up, thumbed through it, and immediately wished
I had found this book ten years ago.
Although written in 1980 and therefore somewhat outdated, this
booklet neatly summarizes a wealth of facts, examples, and references that I
have seen over the years into a concise desk reference. It is clearly meant for
practitioners such as an engineer on the floor of a manufacturing plant. The
booklet is part of a “continuing project” of the American Society for Quality
Control, which is now called the American
Society for Quality (ASQ). The goal of the series is “to survey topics…,
show what sorts of problems they can solve,” and tell where to find “further
techniques” in a “practical (usable) form” (p. ii).
As I thumbed through the booklet, I felt like a Goldilocks
who, at long last, had encountered a reference that was neither too difficult,
not too elementary, but was “just right” for my needs. I was struck by several
aspects of the booklet:
- At 51 pages (not counting appendices), it is
extremely thin. About one-third of the content is theory, and the remainder contains
concrete applications and step-by-step examples
- You can use the booklet to carry out a dozen or
so goodness-of-fit tests, or as a reference from which to research more
sophisticated tests.
- Instead of merely giving the bibliographic
references, each reference is annotated with a sentence that describes the
article. For example, Stephens
(1974) JASA, “describes the use
of the EDF statistics when the parameters must be estimated and compares the
power to other tests for distributional assumptions.” If you are unfamiliar with the literature,
this is an extremely useful feature.
The editor and writers involved in this project put
considerable effort into making sure that this booklet would be as useful as
possible to a wide range of practitioners. I commend the ASQ for providing this
kind of resource for quality engineers back in 1980. Furthermore, the foreword
of the booklet says that the contents were “the subject of a 1.5 hour tutorial
at the 1980 ASQC Annual Technical Conference.”
Fast forward to 2011. Does today’s ASA provide similar
no-nonsense resources for its practitioners?
Does the ASA sponsor short, practical, data-driven courses as part of
its continuing education series? Are the
courses accessible to someone who has been out of school for a while? Do they
present the same tools used in industry? Are the talks and presentations at
national meetings accessible to non-experts?
The ASA is trying to address issues such as these. In
particular:
-
2012
marks the second anniversary of the ASA’s Section for
Statistical Programmers and Analysts (SSPA), which is committed to serving
the needs of practicing statisticians. The SSPA sponsors webinars, JSM sessions,
and continuing education courses that appeal to statistical programmers and
analysts.
-
Next month is the ASA’s inaugural Conference on
Statistical Practice. This conference can help statisticians in business
and industry learn and share best
practices in statistical analysis, programming, and consulting.
Hopefully, many practicing statisticians will attend statistical conferences and workshops in 2012 and beyond that are neither “too hot” nor “too cold,” but are “just right!”