The demonstration of statistical associations between
states or events naturally suggests that the congruity may be more than
coincidental. Possibilities include that the two (or more) states have
an underlying common cause, or that one event or state is affecting or
causing the other. Nonskeptical acceptance of the latter post
hoc, ergo propter hoc theorem may result in erroneous conclusions.
A notable example occurred early in the last century with the
statistical demonstration that individuals who contracted tuberculosis
appeared to have a lower subsequent incidence of bronchogenic
carcinoma. A well-known statistician of that era suggested that the
tuberculous infection somehow conferred some protection from neoplastic
change. An equally logical explanation, which initially eluded the
thinkers of that era, was that the tuberculous disease, for which there
was little effective therapy, carried with it a high mortality that
prevented many younger consumptives from reaching the age levels in
which pulmonary neoplasms have a significant incidence.