An important idea in biomedicine is that some factors can moderate the effects of others. In statistical terms this is called interaction, and it occurs when products of variables are introduced into linear models. Although interactions are clinically important, the statistical models are sometimes misunderstood. In one common misuse, treatment effects are obtained from the wrong place in linear model statistical output, and portrayed in a manner that can reverse their true roles. The issues are laid out here in a straightforward manner to assist alternative medicine researchers in avoiding this kind of mistake and recognizing when it is committed in the literature. It is also emphasized that there are situations in which interactions might be the central issue in a study, and it is illustrated how this might handled.