Stat 2216 Spring 1996
Statistical Methods................................................................ R Claycombe
Second Exam

Write all answers in your blue book and show all work there. Return your exam in your blue book.

20 pts.
1) Inspect the data below.
Year GDP
1 200
2 208
3 217
4 227
5 238
6 240
7 252

a) Estimate (by hand) the best model that you can to predict GDP.
b) Find the point and interval estimate (99%) for GDP in year 7.

20 pts.
2) Suppose that we want to explain herd sizes and all we know is the type of cows in each herd. An assistant compiled the data below not knowing just how we wanted to code things. We can use D = 1 if dairy and = 0 if beef or B = 0 if dairy and = 1 if beef.

Here are the data:
Cows D B
15 0 1
16 0 1
17 0 1
171 0
181 0
19 1 0
a) We start with equation: Number Cows = a + bD .
Find a and b in the regression either by hand or SPSS.
b) Now suppose you think that you might add B to the model. Try to estimate by hand or machine. Explain what happens.

60 pts.
3) Get the gss.sys file and create a printout of all that you do below.
a) Estimate a model where Income = a + b1Age + b2Sex.
b) Interpret b1 for part a and draw a diagram of your model.
c) Now add a dummy variable for marriage, where M = 1 for married respondents and M = 0 otherwise. (If you don't know how to set this up, ask me.) Interpret b1 again and explain why it changes from part a.
d) Find the partial correlation coefficient for M. Does this variable seem to belong in the model? Why?
e) Analyze the residuals of the part c model. Do you find any econometric problems. Explain.
f) Now think about the age variable, does the result you got for its effect make sense? Explain. Reestimate the model so that you allow for a nonlinear age effect. Interpret your results.
g) What statistics can you see in the part f results that substantiate the notion that the part f model fits better (than part c)?

I have neither given nor received unfair aid on this test.