Chapter 23
- Statistical Methods in Anesthesia
- Stanley H. Rosenbaum
Introduction
The advance of scientific knowledge in the past half millennium
has been closely related to the linkage between experimental results and mathematical
assessment of empirical observations. It is the role of the branch of mathematics
termed "statistics" to describe the results of experiments and analyze, as best as
can be done mathematically, the causes and effects behind the experiments.
A serious problem for a clinician or clinical scientist who wishes
to use statistical methodology is that it is a modern branch of mathematics, much
of it developed in the past century. Its mathematical sophistication is often considerably
beyond that of the university-level calculus that is the upper limit of conventional
medical training. Furthermore, many statistical methods involve extremely long and
tedious calculations that require computerized techniques to be practical. This
drawback leads us to the quandary that although statistical methods are useful and
often essential to the understanding of medical science, they may be subtle to comprehend
and difficult to use.
It cannot be the role of a chapter in a book such as this to derive
or even genuinely explain the mathematics behind statistical methods. Many accessible
introductory texts give some of the reasoning and rough, cookbook approaches to practical
usage (see the Selected Readings list). In real life, almost no one ever does any
sort of statistical calculations beyond the most simple with paper and pencil. Application
programs for statistical calculations are widely available for personal computers.
Many of them are reasonably well documented and simple enough to use without much
preparation.
The problem we have with these statistical programs for personal
computers is that the assumptions behind the statistics remain subtle. It is very
easy to use the wrong method or approach and still have the program generate an output
that, though appearing to be correct, is simply wrong. The old computer adage of
"garbage in, (statistically misleading) garbage out" is very easy to achieve with
a computer package and a data file.
It is the goal of this chapter to provide some of the correct
language for describing statistics and to suggest the proper methodology for the
use of statistics in some standard situations. Some of the classic errors common
in medical statistics will be indicated. Of course, readers must be familiar with
whatever program they actually use.
