- Statistics Without Statistics: Did They Use the Right Test?
If you are interested in knowing more of the fine details on statistics, check out
Clinical
Epidemiology: A Basic Science for Clinical Medicine, Sackett et al, Little Brown
(WA 950 S121C 1991, located at Science Library and Grunigen Medical Library). However, if you just
want a sense of what the tests are and how they should be used, read further.
The Variables
After reading the article, formulate the general question the authors are trying to answer. Also note the
type of study used: cohort, case control, randomized, etc. Next, flip through the article. Look at the
tables and graphs. What information is being presented? Are the variables categorical or continuous?
Categorical variables are data sets with definite groups. The variable gender, for
instance, is either male or female. Cholesterol can be broken down into normal, borderline, and high.
Note that categorical variables are usually words and groups.
Continuous variables, on the other
hand, are number sets. A range exists for each variable. Systolic blood pressure, body mass
index, and height are all continuous variables. The categorical cholesterol variable (normal, borderline,
high) mentioned earlier was a continuous variable that was converted into categorical to describe the
population's cholesterol risk. A categorical variable is the descriptive data in an article. A continuous
variable is a set of plotted points that can be drawn through to demonstrate a line. Not necessarily as
descriptive as categorical variables, but generally more powerful in terms of statistical analysis. The
line drawn can demonstrate associations and rates of change. Most of all, the continuous variables in an
appropriate statistical test can display a
normal distribution.
Normal distribution is the crux of all statistics. Most statistical tests assume that the variable being
tested in a population has a normal distribution, where the mean and median of the variable are
approximately equal generating a bell curve. If mean or median deviate from center then the curve can be
skewed from normal. Statistical analyses of skewed variables are less valid.
