### Three-Way ANOVA

```Three-Way ANOVA
Analysis by SAS
Interaction Plot by Excel
The Design
Type of Crime: Burglary or Swindle
Culture of Defendant: American or
Chinese
Sex of Juror: Female or Male
DV: Recommended length of
sentence
Data JurySimulation;
Infile
'C:\Users\Vati\Documents\SimData\
ANOVA3B-13.dat';
Input Culture \$ Crime \$ Sex \$
Sentence;
Omnibus Analysis
 Proc ANOVA;
Class Sex Culture Crime;
Model Sentence =
Sex|Culture|Crime;
 Means Sex|Culture|Crime; run;
Results & Follow-Up
The triple interaction is significant
Conduct simple interaction analysis
Test Culture x Crime for each sex
I asked you to do the simple
interaction analysis even if the triple
interaction was not significant
Simple Interaction Effects
Proc Sort; By Sex;
Proc ANOVA;
Class Culture Crime;
Model Sentence = Culture|Crime;
By Sex; run;
Results & Follow-Up
• For each simple interaction that is
significant, test the simple, simple
main effects.
• Effect of type of crime at given
levels of culture and sex.
• I asked you to do these tests
regardless of whether or not the
simple interaction was significant
Simple, Simple Main Effects
Proc Sort; By Sex Culture;
Proc ANOVA; Class Crime;
Model Sentence = Crime;
By Sex Culture; run;
Compare length of sentence for
burglars versus swindlers when
Juror
Juror
Juror
Juror
is
is
is
is
female, defendant is American
female, defendant is Chinese
male, defendant is American
male, defendant is Chinese
Prepare an Interaction Plot
• Enter the appropriate means into
Excel (2010)
• These are for the female jurors
Insert, Charts, Line
Select “Line With Markers”
Tweak the Vertical Axis
Gee Whiz
Enter Text in Title Box
Then do the same for the Vertical Axis
Enter Chart Title
Other Tweaks
• Resize the chart
• Change fonts
• Remove or change grid lines
• Et cetera
• Then copy the chart to Word,
PowerPoint, or other application.
Prepare the Second Plot
• We need a plot for the male jurors too.
• Enter the appropriate data into Excel.
• Tweak the plot so it has the same dimensions
as the previous plot.
• If you do not, Excel may scale it in a way that
distorts the differences between the two plots.
```