72 Epidemiology of Electromagnetic Fields
Pair- or frequency-matching (e.g., on age, sex, residence, and other personal characteristics)
is common in case–control studies. If the matching variables are true confounders, however,
the resulting study sample is no more representative of the source population, and proper
analytical techniques (matched pair analysis or conditional logistic regression) are needed
to obtain reliable measures of effect. Matching should be used with caution, because the
expected gain in precision may be diminished or nullied by the exclusion of cases without
suitable controls, as well as by unnecessary matching on weak or too many confounders [10].
Confounding can also be controlled in the analysis of data, provid