About the Final

STA431s09: About the Final Exam

Time and Location

The final exam will be on Saturday April 25th, from 12 pm to 3 pm in SE Cafe A.

Office Hours

Jerry will have office hours Wed. April 15th 11-12, Monday April 20th 11-1 and Wednesday April 22nd 11-1.

Christine will have office hours Thursday, Apr 23rd from 2 to 4 and Friday, Apr 24th 3:30 to 5 pm. Note Friday Office Hours have changed!

Don't forget to double check your quiz marks on BlackBoard!

Details about the Exam

The exam has five questions. Two have just one part, one has parts (a) and (b), and two have parts (a) through (h). The questions are on the first five pages of the exam. The last five pages consist of SAS printout (proc calis, of course). A calculator is allowed, but you won't need it.

Model identification appears somewhere in four out of the five questions. This should come as no surprise. Know the identification rules and how to prove them.

Perhaps the best way to study is to read over the class notes and then re-do selected homework problems. I hope the in-class examples will help with the problems.

Here are some more specific suggestions about the assignments.

Review material will not be on the exam directly, so don't bother with Assignment 1. The materiaal on Assignment 2 is also not on the exam, except as you might need it to be able to do calculations on structural equation models.
The multivariate normal likelihood is important. Please review Question 1 of Assigment 3.
Assignment 4 is mostly review, except for Question 6. Re-do 6a and let the rest go.
Assignments 5 through 11 are where the action really is. If you can do these problems, you should do well on the final. For the SAS parts, just make sure you understand what is in your output.

The SAS part of the exam will be based on some data obtained from the United Nations. For 97 countries we have birth rates, death rates, infant mortality, life expectancies for males and females, and Gross National Product -- all as reported by the governments. The variables (with my SAS variable names) include:

Live birth rate per 1,000 of population (birthrat)
Death rate per 1,000 of population (deathrat)
Infant deaths per 1,000 of population under 1 year old (infmort)
Life expectancy at birth for males (lifexM)
Life expectancy at birth for females (lifexF)
Gross National Product per capita in U.S. dollars (gnp)

You will not write any SAS code. You will look at my code, and answer questions about the model and about the output. The best way to prepare is to think about this example. Try to come up with a reasonable, simple model. Make a path diagram. Is your model identified? Prove it. Try more than one model.

The rules are that Christine and I will not tell you what the "right" model is, even if you come up with exactly what is on the exam. If asked, we will look at your model, listen to your ideas, and ask you questions. One hint: Of course there must be a latent variable, but you knew that already.