James M. Sherlock in Psychology Today:
Let me introduce the problem with an example. In a recent article from The Conversation, researchers from the University of Queensland describe a study in which they observed families during hospital visits and measured the anxiety and distress of both parents and children. Measures of children’s anxiety and pain were positively associated with parents’ distress levels. The researchers interpret the role of the parents’ distress during the visit as causal, arguing that parents who are more anxious and distressed are less able to support their child during the procedure. This seems like a rather sensible and intuitive conclusion to draw; however, the researchers should have absolutely no confidence in doing so.
Anxiety sensitivity (i.e. how susceptible an individual is to feeling anxious, ranging from completely normal to pathologically desensitise/hypersensitive) is substantially genetic. In fact, nearly half of the variation (i.e. the differences between people in the population) in anxiety sensitivity is likely to be genetic. The researchers therefore have no way to distinguish whether the children are responding to their parents’ distress or simply share their anxious disposition for genetic reasons, or more likely some combination of the two.
More here.