Suppose that a researcher wants to study the effects of chocolate consumption on attitudes about the nutritional effects of candy. The researcher believes that people will have more favorable attitudes about the nutritional benefits of candy after consuming a chocolate bar. The researcher selects two groups of subjects (Group C and Group NC), each consisting of 20 students. The subjects in Group C are recruited from students in a History of Chocolate course. The subjects in Group NC are recruited from an Introduction to Bowling course. The subjects in Group C will eat a chocolate bar and the students in Group NC will eat no chocolate. Then all subjects will be given a survey that measures attitudes about the Nutritional Benefits of Candy (the NBC survey), which yields a score between 0-100. A high score on the survey indicates that the subject believes that candy has large nutritional benefits. The researcher plans on (1) conducting the research in his lab, (2) allowing subjects in Group C to choose the brand of chocolate bar they will eat from two options, and (3) testing the null hypothesis at the .01 level of significance, but the researcher has flexibility in these aspects of the research design. However, the researcher does not have access to any additional subjects.

However, the researcher changes his design by going international. He chooses two additional groups of subjects from the University B (International), one (from a History of Chocolate course) that will eat chocolate and one (from an Introduction to Bowling course) that will not eat chocolate. Then both groups take the NBC survey. The researcher performs a two-way ANOVA using the University A (Domestic) data and the University B (International) data and finds a statistically significant interaction.

1. What does the interaction mean in the context of this research? (Refer to chocolate consumption, attitudes about the nutritional benefits of candy, and the university that a subject attends.)

2. What do the results of the ANOVA tell us about the external validity of the results if the researcher had used data from only Domestic students?