Preview File Edit View Go Tools Window Help Chegg.com a Final 2017 (page 1 of 2)

Question

Preview File Edit View Go Tools Window Help Chegg.com a Final 2017 (page 1 of 2) bbeswebda.+ a Search 2) The effect of attendance on performance in economics principles exams has traditionally been investigated by using attendance as an explanatory variable in addition to traditional explanatory variables such as SAT, GPA, gender, and age. One problem with tis is that one never knows if the negative relationship traditionally uncovered is because poor students are more likely to miss class, or if missing class causes a lower exam score. A journal submis- sion has addressed this problem via a probit analysis using data from 95 students on a 30-question multiple-choice exam. The dependent variable is whether or not a question was answered correctly. The explanatory variables were a set of 94 dummies for the students, a set of 29 dummies for the questions, and a dummy for whether or not the student was absent whenever the course content relating to that question was presented in class. A referee says this is crazy because it doesn't control for whether a student is smarter SAT score), works harder (GPA), is male, is older, etc., and that there are far too many degrees of freedom lost by a those dummies. The journal editor doesn't understand fancy econometrics and so comes to you for advice. What advice would you provide? POST GUESTION 19 questions remaining this subscription period 6

Explanation / Answer

I agree with the referee. The 94 dummies for students and 29 dummies for questions is pointless. When he put 94 dummies for students, he’s trying to control for variation in responses due to student’s individual trait. However, it is true that we lose too many degrees of freedom. However, 29 dummies for questions is not needed at all since all students were exposed to same set of questions. Before adding too many dummy variables, we must check if a control belongs to a model or not. Individual level student variation captured by dummy variable won’t be any useful if we wish to use the model for prediction purposes. This is because it is largely unobserved variation. We cannot observe how student A differs from student B. It is certainly better if we use the observables such GPA, SAT score, Age and gender dummy and build relevant model.

Related Questions

Navigate

• Browse (All)
• Browse P
• Subjects

---

---

Integrity-first tutoring: explanations and feedback only — we do not complete graded work. Learn more.