STAT 529
Lecture 7
Student Notes
Assignment #1 Due on Tuesday in Lecture
Info on Cholesterol Data (Q. 5 Assignment 1)
Compare women and men HDL data
130 observations women & 119 observations men
Unknown difference between women and men
1. States of Nature
STAT 529  Lecture Note 1
2015
Jan 13,
Course Overview
3 Assignments 25% each
Final take home exam 25%
Syllabus, assignments and lecture notes will be uploaded in the course website.
Class assignment: A randomized response
What is a sensitive question as
Stat 529: Spring 2015
Lecture 21 (03/31/15)
1.
Re: A3  Sigmas for the lamb data (A3 part 1) are in the lecture notes.
Part2 of Assignment 3 is up on the website.
Assignment 3 is due April 14th.
2.
Discussion of A3 Q3
Corn data:
F1
C1
Cell 1
X11
= [X111,
Lecture 19
Assignment 3, Part 1 is on the web, the second part will be uploaded next week.
SUGGESTIONS MADE FOR ASSIGNMENT SUBMISSION DATES
April 14  A3 to be turned in
special session on Friday by Prof. Deely to clear any last minute doubts
May4 day to
Lecture16 STAT529
Note: For assignment 1, a score of 7085 is an A while 5570 is B.
This lecture was mainly about question 5 of assignment #2. The question uses the Lamb data.
There are 5 different diets for Lambs for a period of 2 weeks.
The frequenti
Lecture 17
STAT529
03/10/2015
Section1: Discuss about Assignment2, Q4 (b)
In the original question, it requires two empirical methods:
1. toothpicks
2. moment method
In the moment method:
f ( x )= f ( x ) ( ) d
E ( x )=
Var ( x )= (1+ )
x
converge
S 2
STAT 529 Lecture 15 Class Notes
Review of statistical notations:
f (x, y) is the joint distribution of random variable x and y;
We have f (xy) g(y) = f (x, y);
Then
E [ x ]= xf ( x , y ) dydx= xf ( x y ) g ( y )dxdy = cfw_ xf ( x , y ) dx g ( y ) dy= E
LECTURE 18
Suppose you have a function
( )
from which you want to simulate
values, but you cannot recognize it well enough to do it.
The MetropolisHastings algorithm is a Markov chain Monte
Carlo (MCMC) method for obtaining a sequence of random samples f
Lecture 12
Noninformative Priors
There is not uniform agreement about the uniqueness of the noninformative prior
to choose
1. Laplaces prior
uniform
2. Improper prior, such as
1
1
3. Jeffreys prior (today topic) (improper)
4. Zellners prior
whic
Lecture 14 (2/26/2015) Note taker: WenYu Chao
Assignment 2 Question 4
The data X1,X50 will be later put on the web.
Assume the prior is a gamma distribution with the parameter and.
1
1
i.e. ( , )= ( ) e
This assumption also allows us to work on the fol
Session 10
In this session, we have further work in how to make decision model cover many problems.
Interval Action Space
(, ) is OK.
: Some interval in real line, i.e.,
A :cfw_[ a , b ] :a ba ,b
This is extending the estimation situation.
cfw_
1 , if [
Stat 529 Note 1
By Tong YU 01/13/2015
sensitive questions?
Information for secret folder
ID: stat529
Password:
Decision model:
= R the state of nature
>get an operational definition of the states of nature
e.g. the mean height of all male student at Purd
Lecture 22 (04/02/2015)
For HW#2, A: 65+ ; B:44+ ; C:35+
Treatment:
New
T
Usual
S
We would say that
T
s
is better by 30 than
when:
P ( T s 30 ) 0.9
2
T , s N ( , 2 ) , 2
For noninformate:
n
2
(
( T x n) =N x n ,
n
So:
)
2
(
( s y n )=N y n ,
n
)
G
STAT 529 03/26/2015 Lec 20
Comments on A3Q1:
In method 2, there is something you can try:
1. Take the hyper prior distribution of alpha and beta.
2. Generate thousands of betas and alphas
3. Make a histogram of thetas.
The histogram is the prior distribut
STAT 529  Lecture Note 4
2015
Jan 22,
Class assignment: A randomized response
Instructions for Randomized Response Collection is updated on course website
Data (yes/no two responses) should be given on 1/29/15.
Randomized Experiment:
If yes,
Y i=1,
(or)
Stat 529: Spring 2015
Class Notes: 1/27/2015 (Lecture 5)
1. Instructions for the Randomized Response Collection experiment
We have to ask two people: one Domestic student and one International student to
participate in the experiment and we need to submit
01/29/2015 Thursday
Announcements:
HW1 will due next Thursday. Hard copy in class.
Homework is shorter. There is a new version posted.

C scan data:
Freq risk:
x1
r ( t , 1 ) = L ( t ( x i ) ,1 ) f ( xi 1) =L ( t ( , 1 ) f ( x 11 ) ) + L ( t ( x 2)
Stat529 Class Note 2
1. Interesting sensitive question review
Candidate questions: 1) Have you used illegal drugs? 2) Have you cheated in previous exam? 3)
Are you current using illegal downloaded or stolen software or materials? 4) Do you think the
curr
Stat 529: Spring 2015
Class Notes: 1/20/2015 (Lecture 3)
1. Homework Assignment 1 is up

Assignment will be in two parts. Part 1 is up
Assignment will be due in two weeks. This will place the due date between 2/3/2015 and
2/10/2015
Assignment details wil
Lecture notes on Apr. 14, 2015
There are four sets of homework. Each takes 25% of the final grade.
One may get an A if he/she obtains more than 80% of the overall points.
If one gets 70% in HWK1 and 80% in HWK2, the professor will give 75%
for HWK1.
Backw
STAT 529  Lecture Note 2
2015
Jan 15,
Discuss RSS article show on overhead and Steve the Bayesian
This made it to the top of the "most emailed" list last night.
http:/www.nytimes.com/2014/09/30/science/theoddscontinuallyupdated.html?
action=click&pgt
Lecture Note 24 Stat 529 Spring 2015
Correction for A3 extra credit question (delete b) and c),
compute posterior probability only. Be sure to look at A3 Part 2
as it is modified.
Students who choose to do the extra credit question, please turn
in extra t
STAT 529 04/07/2015 Lec 23
AnnouncementA3Q4 Extra Credit
A3Q4: Some hints
Choose a few value of b by yourself. Make b practical significant.
You have to decide hyperprior.
A3Q5
Part 1
t , s , n
t
we will say
s
is better than
t s +30
if
, n will be 200 in
Stat529 Class Note 2
1. Interesting sensitive question review
Candidate questions: 1) Have you used illegal drugs? 2) Have you cheated in previous exam? 3)
Are you current using illegal downloaded or stolen software or materials? 4) Do you think the
curr
Lecture26
This lecture is given by Professor Wisely Dunson.
This lecture is about Logistic Regression.
The failure of an Oring seal was determined to be the cause of the Space
Shuttle Challenger disaster on January 28, 1986. A contributing factor was col
Stat 529  Applied Bayesian Decision Theory
Lecture 11  02/17/2015
By Joonsoo Kim
A. Prior based on histogram
Let's find generalized beta CDF using histogram of observed data.
1. Compute histogram of observed data, and it can be made into a pdf by multip