Nuisance Variable

A nuisance variable is an unwanted variable that is typically correlated with the hypothesized independent variable within an experimental study but is typically of no interest to the researcher. It might be a characteristic of the participants under study or any unintended influence on an experimental manipulation. A nuisance variable causes greater variability in the distribution of a sample’s scores and affects all the groups measured within a given sample (e.g., treatment and control).

Whereas the distribution of scores changes because of the nuisance variable, the location (or the measure of central tendency) of the distribution remains the same, even when a nuisance variable is present. Figure 1 is an example using participants’ scores on a measure that assesses accuracy in responding to simple math problems. Note that the mean is 5.00 for each of the two samples, yet the distributions are actually different. The range for the group without the nuisance variable is 4, with a standard deviation of 1.19. However, the range for the group with the nuisance variable is 8, with a standard deviation of 2.05.

For the first distribution, participants’ math accuracy scores are relatively similar and cluster around the mean of the distribution. There are fewer very high scores and fewer very low scores when compared with the distribution in which the nuisance variable is operating. Within the distribution with the nuisance variable, a wider spread is observed, and there are fewer scores at the mean than in the distribution without the nuisance variable.

Figure 1 Frequency Distributions With and Without Nuisance Variable

Notes: (a) Without nuisance variable (N = 240). (b) With nuisance variable (N = 240).

[Page 1098]In an experimental study, a nuisance variable affects within-group differences for both the treatment group and the control group. When a nuisance variable is present, the spread of scores for each group increases, which makes it more difficult to observe effects that might be attributed to the independent variable (i.e., the treatment effect). When there is greater spread within the distributions of the treatment group and the control group, there is more overlap between the two. This makes the differences between the two groups less clear and distinct.

Figure 2 is an example using treatment and control groups, in which the independent variable is an extracurricular math tutoring program and is received only by the treatment group. Both samples are tested after the administration of the math tutoring program on measures of math accuracy. In this case, a nuisance variable might be the participant’s level of anxiety, but it could just as easily be an external characteristic of the experiment, such as the amount of superfluous noise in the room where the measure of math accuracy is being administered. In considering the effects of the nuisance variable, the distributions of the two groups within the sample might look similar to that in Figure 2.

In the distribution without the nuisance variable, the variation in participants’ anxiety is reduced or eliminated, and it is clear that the differences in participants’ observed math scores are, more than likely, caused by the manipulation of the independent variable only (i.e., the administration of a math tutorial to the treatment group but not the control group).

...