Factorial Designs

Step 2: Plot your data

Begin by selecting Estimated Margianl Means below the ANOVA dialog.

In the dialog area tjat opens, you can see places to drag your independent variables from the left to the right. Drag Attractiveness first and then Commitment: It should look like below. Select Marginal Mean Table under Output and Observed Scores under Plot.

The y-axis will always be a dependent variable, and a dependent variable is a response you get back from participants. Which of the three variables best fits that description? Two of the three variables are factors controlled by the experimenter, and only one is a response provided by the participants. Drag that variable onto the y-axis.

The two independent variables will go along the x-axis or into the “Cluster on” box (the legend). The choice of which independent variable goes where is more art than science. Sometimes, the story is easier to tell with one variable in the legend and the other along the x-axis. If you find it hard to interpret the data with the plot done one way, try the other way. Go ahead and put Attractiveness along the x-axis and Commitment in the legend. Press OK to make the plot. You should get this:

Take a moment to look at this plot. First, look at the gold cirlces and lines. Those represent the single (“Low Commit”) participants. They show the expected difference in ratings between High-Attractive and Low-Attractive partners. What about the blue circles and lines? Those represent the committed participants. They do not show the same pattern. Compared to Low-Commitment subjects, the High-Commitment subjects show much lower ratings of the High-Attractive partner. These results support the researchers’ hypothesis.

The plot above has a number of advantages:

  1. By plotting the means, you get a clear picture of the pattern: Attractiveness has the expected effect on Low-Commitment subjects, but High-Commitment subjects seem to be devaluing the High-Attractiveness partner.
  2. By plotting confidence intervals, you can see the precision of the estimates for the means. In this case, the confidence intervals are narrow, indicating relatively high precision.