This exercise is designed to show differences in findings depending on whether you use simultaneous or sequential regression, and depending on the order of entry in sequential regression. Angela Duckworth and Martin Seligman (2005) were interested in the effects of self-discipline on students’ academic performance. They measured the self-discipline

(degree of self-regulation and lack of impulsiveness across multiple domains) of 154 eighth-graders in the Fall, and used that to predict (or was it explain?) final GPA in the Spring. Also controlled were students’ IQs and their previous GPA, also measured in the Fall. The data are in the file “Duckworth Seligman sim data.sav”; the data are simulated but are designed to produce results consistent with the findings of the original study.

a. Conduct a simultaneous regression of GPA on IQ, previous GPA (Pre_GPA), and self-discipline (Self). Which variables are important in explaining GPA? In particular, how important are IQ and self-discipline?

b. Conduct a sequential regression of GPA on these same variables. For this regression, enter IQ, Self, and Pre_GPA, in that order. Draw the causal model implied by this regression. Using DR2 and the b for each variable (as it is entered), note the relative importance of the variables. Again, focus in particular on IQ and self-discipline.

c. Conduct another sequential regression on these same variables. This time the order of entry should be IQ, Pre_GPA, and Self. Draw the causal model implied by this regression. Compare these results to the results from step

b, with particular attention to the self-discipline variable.

d. Explain the reason for the differences in the “importance” of the variables from the different regressions.