Blackwell Reference Online

Extract

Control variables are variables included in multivariate analyses to identify spurious associations. In assessing whether X is associated with Y , it is important to examine whether the covariation between them persists after the effects of other variables on this association are removed ( McClendon 2002 ). Here, “control” means to hold constant. A variable is controlled when its influence on the other variables in the model is held constant. In laboratory experiments a variable Z can be controlled by setting a fixed value for it and observing the relationships between X and Y for that fixed value. The experiment can then be repeated at other fixed values of Z to see whether the same results occur. However, in most social science research, values of variables such as race, education, age, and income cannot usually be manipulated before obtaining the data. In correlational studies the portion of the association between X and Y that is caused by variation in Z can be removed by comparing only cases with equal or similar values of Z at a time. A spurious relationship exists between X and Y if both variables are dependent on Z , so that the association between X and Y disappears when Z is controlled (see figure 1 ). Figure 1 Graphical representation of a spurious relationship between X and Y Informed and methodological application of control variables ... log in or subscribe to read full text