Expands the Capabilities of SPSS and Excel 

Program Dynamic

When you do a factorial analysis of variance, it is rare that you look only at the overall effects and stop there.  Rather, you typically conduct more focused tests, called single degree of freedom contrasts.  These contrasts might focus on the comparison of two means from a main effect, two means from a simple main effect, or an interaction contrast that compares the difference between mean differences (for a definition of these types of contrasts, Click Here).

The GLM routine in SPSS provides you with single degree of freedom comparisons for main effects.  If you know how to use the syntax features of the program, you also can have the program do single degree of freedom simple main effects.  To get interaction contrasts, you must use the /LMatrix command in program syntax, which is cumbersome. 

ZumaStat has a utility that automatically calculates all of the single degree of freedom contrasts that a researcher will typically be interested in.  Input is the cells means and sample sizes that are read from an ascii file.  In addition, you provide the mean square error from the analysis of variance summary table (that might be generated using SPSS) and its degrees of freedom.  With this information, ZumaStat generates the contrasts.

This input information typically is available in journal reports, so you can reanalyze data from published studies using ZumaStat.

ZumaStat uses a moderator variable framework to organize the contrasts. The framework makes distinctions between outcome variables, independent variables and moderator variables.  The outcome variable is the dependent variable in the analysis. The independent variable is the presumed cause of the dependent variable and one of the factors is designated the independent variable.  In the case of an interaction, the effect of the independent variable on the dependent variable changes depending on the value of a third variable, called the moderator variable.  The other factor in the design is designated as the moderator variable. 

In the example below, the outcome variable is relationship satisfaction with one's parent and the two factors are grade in school and gender.  The researcher decides to let gender be conceptualized as the independent variable and grade as the moderator variable.  Thus, the interest is in gender differences in relationship satisfaction and how these might vary as a function of grade.  ZumaStat's Help menus have extended discussions of this framework.

ZumaStat generates single degree of freedom contrasts for the main effects, for the effects of the independent variable at each level of the moderator variable (called simple main effects) and for the relevant interaction contrasts.

ZumaStat also has a routine for three factor designs.  It also has routines for factorial analyses when the assumption of homogeneity of variance in the population is untenable.  Finally, for large designs and analysis of covariance, ZumaStat has an automatic syntax generator for the /LMatrix command in SPSS to allow you to cut and paste syntax for complex interaction contrasts.

The output is shown in an ascii window and is discussed below.

How it Appears on Your Screen

Output

Here is what the output looks like:

The output first shows the input data and then reports the relevant single degree of freedom contrasts.  For a given contrast, you get the parameter estimate for the contrast (i.e., the value of the contrast), its estimated standard error, the confidence interval, the t ratio for the significance test of the contrast, the p value for the t ratio, and the partial eta squared for the contrast.

In the above example, ZumaStat generated 10 single degree of freedom contrasts.  In the analyses, there were gender differences at each grade level (as indicated by statistically significant contrasts C5, C6, and C7).  However, the gender difference at grade 7 was not statistically significantly different from the gender difference at grade 8 (contrast C8) nor grade 9 (contrast C9).  Similarly, the gender difference at grade 8 was not significantly different from the gender difference at grade 9 (contrast 10).

ZumaStat makes it easy to do theoretically guided single degree of freedom contrasts.   

To control for experimentwise error rates across multiple contrasts, you can use the Holm modified Bonferroni procedure offered by ZumaStat,