Confounding variables must be controlled for in order to draw valid causal conclusions.

The confounding effect of socioeconomic status made interpretation of the results problematic.