Stats 300C: Theory of Statistics
Spring 2013
Lecture 2 April 3, 2013
Prof. Emmanuel Candes
Scribe: Will Fithian (2012), slightly edited by E. Candes (2013)
1
Outline
Agenda:
Global testing
1. Needle in a Haystack Problem
2. Bonferronis Global Test
3. Opti
Hashtag network
The hashtag network is undirected. After export data from SQLite, there are two
columns and more than three hundred thousand rows. Since the number of
hashtag is far larger than tweet id, so there exists huge invalid tweet id. Before
graph
BIA654AWS Experimental Design
Chapters 9, 10, 11, & 13 and Cumulative
Prof. Thomas F. Brantle, Ph.D.
Final Exam Brief Solutions
Final Exam Chapters 9, 10, 11, & 13 and Cumulative Brief Solutions
This is an open book and open notes exam. You are expected t
BIA654AWO Experimental Design
Cumulative
Prof. Thomas F. Brantle, Ph.D.
Final Exam
Final Exam Cumulative
This is an open book and open notes exam. You are expected to work independently and not
share information. The point value for each question is clear
#include<iostream>
#include<string>
#include<vector>
#include<fstream>
using namespace std;
class text
cfw_
public:
unsigned long long int num;
string name;
text()cfw_ num = 0; name = "\0";
;
class text1
cfw_
public:
unsigned long long int num;
text1()cf
#include<iostream>
#include<string>
#include<vector>
#include<fstream>
using namespace std;
class text
cfw_
public:
unsigned long long int num;
string name;
text()cfw_ num = 0; name = "\0";
;
class text1
cfw_
public:
string name1;
string name2;
text1()cf
Stats 300C
Homework 1 Solutions
Problem 1
Part (a) In order to explore the eect of positive correlations on the probability of Type I Error
when using Simes procedure, we generated test statistics from the equicorrelated multivariate normal
model given by
Stats 300C
Homework 4 Solutions
Problem 2
Let BS = cfw_ Rp | i = 0 i S . Up to constants which do not depend on S , P(S |y ) P (S, y )
/
S P(y |S ), so we begin by deriving an expression for P(y |S ).
We begin by noting that because satises the normal eq
Stats 300C
Homework 3 Solutions
Problem 1
To show that extended Simes test is more liberal than the Hochberg procedure, it suces to show
that the closed Simes test can reject even when the Hochberg procedure accepts.
To this end, with = 0.05, take
p(1) =
Problem Set 2 Solutions
April 27, 2013
1
Of course, if we make no adjustment for multiple comparisons, then any experiment making many comparisons is quite likely to have some false rejections.
It is common in modern studies to test hundreds of thousands
Stats 300C
Homework 5 Solutions
Problem 1
There were a number of really good solutions to this problem. We compare the risk of hardthresholding, soft-thresholding, BH estimation, and soft-BH estimation. The latter procedure is
performed by using BH to est
Stats 300C: Theory of Statistics
Spring 2013
Lecture 1 April 1, 2013
Prof. Emmanuel Candes
Scribe: Will Fithian (2012), slightly edited by E. Candes (2013)
1
Outline
In this lecture, we cover the following topics:
1. Multiple Testing
2. Needle in a Haysta
Stats 300C: Theory of Statistics
Spring 2013
Lecture 5 and 6 April 10-12, 2013
Prof. Emmanuel Candes
Scribe: Will Fithian
1
Outline
Agenda:
Global testing
1. Tukeys Higher Criticism
2. Detecting Sparse Heterogeneous Mixtures
3. Threshold Eect
Last Time:
W
Stats 300C: Theory of Statistics
Spring 2013
Lecture 4 April 8, 2012
Prof. Emmanuel Candes
Scribe: Will Fithian
1
Outline
Agenda:
Global testing
1. Fishers Combination Test
2. Simes Test
3. Tukeys Second-Level Signicance Testing
Last Week: We studied two
Stats 300C: Theory of Statistics
Spring 2013
Lecture 3 April 5, 2013
Prof. Emmanuel Candes
Scribe: Will Fithian (2012), slightly edited by E. Candes (2013)
1
Outline
Agenda:
Global testing
1. 2 test
2. Detection Thresholds for Small Distributed Eects
Glob