Vietnam Quantitative Finance Society

User Name
Password
Closed Thread
09-21-2007, 06:13 AM   #11
Khoa Tran

Core Member
Master Sergeant
 
Khoa Tran's Avatar
 
Join Date: Jun 2007
Posts: 211
Thanks: 28
Thanked 19 Times in 16 Posts




Default

Regarding martingales, I have another fun problem. If the Monty Hall Problem is too popular, this one can provide more fun for an interview.

23. (Likelihood Ratio Test paradox?) a. Let \{X_n\} be a sequence of independent & identically distributed (iid. for short) random variables. You don't know it's true distribution but suppose you know that the probability density function is either p(x) or q(x) (if x is a vector then p and q are joint density functions). In statistics, the ratio
\lambda_n=\frac{q(X_1,X_2,...,X_n)}{p(X_1,X_2,...,  X_n)}
is known as a likelihood ratio. After each experiment, you update your \lambda_n. The ratio \lambda_n is likely to be small or large, according as the tru probability density function is p or q, and is used as a statistical test to determine the true distribution.

However, show that if p is the true density, \lambda_n is a martingale.

b. If it is a martingale then why is it used as a statistical test? What's wrong?

You can run a simple simulation here (Matlab code) to see what happens.
Code:
%A deformed coin with 70% head and 30% tail
%You know that it's deformed
%You guess that it's either: (70% head, 30% tail) or (30%head, 70% tail)
%Now run the likelihood ratio test for 100 times
ratio=1;
for i=1:100
    r = rand;
    if r<0.7 %if head
        ratio = ratio*0.3/0.7;
    else %if tail
        ratio = ratio*0.7/0.3;
    end
end
ratio
You will "always" get almost 0 (say, of order 10^{-10}).
Khoa Tran is offline
10-18-2007, 09:47 AM   #12
Khoa Tran

Core Member
Master Sergeant
 
Khoa Tran's Avatar
 
Join Date: Jun 2007
Posts: 211
Thanks: 28
Thanked 19 Times in 16 Posts




Default VNQF brainteasers collection

Có lẽ sau chiêu sư tử hống n\sin(n) (link) của Kim Mao Sư Vương Ngô Quang Hưng, quần hùng lăn ra giãy đành đạch, không còn ai hứng thú thi triển võ công nữa. Thôi thì chúng ta lại viết tập mới với mấy câu hỏi dễ thương hơn.

1. Let W_t be a Brownian motion. What is \mathbb{E}[W_t^6]?

2. How do you compute \int_0^\infty e^{-3x^2}dx

3. (from YADD) After each second, a bacteria may: either die, or survive, or split into 2, or split into 3 (each with probability 0.25). Suppose that there's 1 insect at the beginning of the day, what is the probability that the whole population dies at the end of the day?

4. (Programming) Given 2 variables a and b. How do you swap the values of a and b without using any extra memory?

5. (Blog KHMT) Giả sử

(mũ x đến vô cùng). Vậy x bằng bao nhiêu? Ví dụ, ta có thể lý luận như sau. Dễ thấy rằng . Do đó , hay . Nhưng nếu thì, theo lý luận trên, cũng dẫn đến . Sao lạ thế? à?

Last edited by Khoa Tran; 11-17-2007 at 11:14 PM.
Khoa Tran is offline
11-13-2007, 06:39 AM   #13
shinichi9htv

Member
Sergeant
No Avatar
 
Join Date: Jul 2007
Posts: 148
Thanks: 13
Thanked 7 Times in 6 Posts




Default

[***] Có 124 con bài khác nhau và 2 nhà toán học. 1 người thứ 3 cầm lấy bộ bài, lấy ra 5 con bất kì, đưa cho 1 nhà toán học. Ông này ngồi xem xem tính toàn một thôi một hồi, đút lại 1 con vào bộ bài và xếp 4 con ngay ngắn trên bàn. Nhà toán học kia đứng lên nhìn 4 con bài, rồi tìm ra con bài thứ 5 còn lại. Hỏi 2 nhà toán học đã làm thế nào?

(Restated by Khoa)

Last edited by Khoa Tran; 11-17-2007 at 11:18 PM.
shinichi9htv is offline
Closed Thread


Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Report on VNQF and VEF's joint meeting in California Khoa Tran General 5 04-08-2008 06:23 PM
VNQF 1st Union in California Khoa Tran General 16 10-23-2007 02:41 AM
Tổng quan VNQF / VNQF Overview Admin General 2 08-12-2007 03:38 PM

Save or bookmark this site with: Del.icio.usStumble It!AddThis.com
All times are GMT +7. The time now is 03:37 AM.
Powered by vBulletin Version 3.7.2