The issues we have considered so far pertain to what is sometimes
called the internal validity of a study or experiment. They
pertain to the conclusions we can draw about cause and effect with
regard to the class of things actually observed or included in an
experiment. Of course, we normally want to generalize from that
class to a wider population. We want to know whether smoking
causes cancer for people in general, not just for those who
happened to be included in a particular study.
This raises questions of external validity, questions you should
ask when a researcher claims that his or her findings apply to a
population as a whole.
To evaluate external validity, we should remember the special
importance of definitions in statistical reasoning. To establish
a correlation between two variables, we must define the
variables in such a way that they can be counted or measured. In
some cases this is fairly easy; in others it is difficult if not