13.3 Running Two-way ANOVA

Open your ezdata.sav file and select Analyze, General Linear Model, Univariate... from the menu (Figure 13.1).

A Univariate dialog window will appear (Figure 13.2). In the left pane, scroll all the way down and select masctot (overall masculinity score), then move it to the Dependent Variable: box on the right. Next, in the pane on the left, select Employee Gender and Masulinity Level (masc) and move them to the Fixed Factor(s): pane on the right (recall that SPSS uses the term, factor, to refer to the independent variable, so these terms are synonymous). This will generate the ANOVA table when we run the analysis. In the event that the interaction effect is significant, we will want to generate a figure depicting this to aid our interpretation. To do this, click the Plots... button on the right.

A Univariate: Profile Plots window will appear (Figure 13.3). This dialog window wants to know which factor to represent on the horizontal axis and which one to represent by drawing two lines. While technically it doesn't matter which factor is placed where, using separate lines for gender will faciliate comparisons between the sexes. Thus, move gender to the Separate Lines: box and move masc to the Horizontal Axis: button.

Next, click the Add button at the bottom of this window. You will now see masc*gender has been added to the Plots: pane at the bottom of this window (Figure 13.4).

Last, click the Continue button in the upper right corner. This window will close, returning you to the main Univariate dialog window (Figure 13.2). The last thing we need to do is to specify our options. Click the Options button on the right of the Univariate dialog window. A Univariate: Options window will appear (Figure 13.5).

First we need to instruct SPSS to calculate the marginal means for the main effects and cell means for the interaction effect. Note that an estimate of these means could be generated by moving the factors and the interaction from the upper left pane to the Display Means for: pane on the right. However, these means will only be accurate if there is an equal sample size in all four cells. This is not the case for our data file.

Further, researchers are also frequently interested in other descriptive statistics (e.g., the standard deviations), and this option only generates means. Thus, instead of using the Display Means option, we will use another way to generate the means (and other descriptive statistics) - the Descriptive statistics option. To do this, check the Descriptive statistics box in the lower left of this dialog window.

In addition to determining whether or not the independent variables yield significant differences in the dependent variable (which will be assessed by the F-values), researchers typically also want to know how big the differences are. To do this, check the Estimates of effect size box also. Last, click the Continue button at the bottom of this window. This will return you to the main Univariate dialog window (Figure 13.2). We are ready to run our ANOVA now, so click the OK button in the lower left of the this window.