In a scientific experiment measuring the effect of one or more independent variables on a dependent variable, controlling for a variable is a method to reduce the confounding effect of variations in a third variable that may also affect the value of the dependent variable. 转自wikipedia