Scribd
Upload
50 views

Stat 100a: Introduction To Probability

This document outlines the topics covered in a probability and statistics course, including: homework assignments, moment generating functions of uniform distributions, correlation, bivariate normal distributions, conditional expectation, and the law of large numbers. It discusses key concepts such as how correlation quantifies the relationship between two random variables, how to find conditional expectations, and how the law of large numbers implies that short-term variations in luck become negligible for average outcomes over the long run.

Uploaded by

Jun Kai Ong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
50 views

Stat 100a: Introduction To Probability

This document outlines the topics covered in a probability and statistics course, including: homework assignments, moment generating functions of uniform distributions, correlation, bivariate normal distributions, conditional expectation, and the law of large numbers. It discusses key concepts such as how correlation quantifies the relationship between two random variables, how to find conditional expectations, and how the law of large numbers implies that short-term variations in luck become negligible for average outcomes over the long run.

Uploaded by

Jun Kai Ong
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 9

Stat 100a: Introduction to Probability.

Outline for the day:


1. HW3.
2. Moment generating function of uniform.
3. Correlation.
4. Bivariate normal.
5. Conditional expectation.
6. LLN.

We are skipping 6.7 and the bulk of 6.3 about optimal play with
uniform hands.
1. HW3. Due Tue Dec 5 in the beginning of class. 6.14, 7.2, 7.8. On 6.14,
assume t < l.

2. Moment generating function of a uniform random variable.

If X is uniform(a,b), then it has density f(x) = 1/(b-a) between a and b,


and f(x) = 0 for all other x.
X(t) = E(etX)
= ab etx f(x) dx
= ab etx 1/(b-a) dx
= 1/(b-a) ab etx dx
= 1/(b-a) etx/t]ab dx
= (etb eta)/[t(b-a)].
3. Correlation.
For any random variables X and Y, recall
var(X+Y) = var(X) + var(Y) + 2cov(X,Y).
cov(X,Y) = E(XY) E(X)E(Y) is the covariance between X and Y,
cor(X,Y) = cov(X,Y) / [SD(X) SD(Y)] is the correlation bet. X and Y.

For any real numbers a and b, E(aX + b) = aE(X) + b, and


cov(aX + b,Y) = a cov(X,Y).
var(aX+b) = cov(aX+b, aX+b) = a2var(X).
No such simple statement is true for correlation.

If r = cor(X,Y), we always have -1 r 1.


r = -1 iff. the points (X,Y) all fall exactly on a line sloping downward, and
r = 1 iff. the points (X,Y) all fall exactly on a line sloping upward.
r = 0 means the best fitting line to (X,Y) is horizontal.
r=0 r = 0.44 r = -0.44.
0.6

6
5
0.2

4
3
eps

y
y

2
-0.2

2
1
0
-0.6

0
-1

-3 -2 -1 0 1 2 3
4. Bivariate normal.
X ~ N(0,1) means X is normal with mean 0 and variance 1.
If X ~ N(0,1) and Y = a + bX, then Y is normal with mean a and variance b2.

Suppose X is normal, and Y|X is normal. Then (X,Y) are bivariate normal.

For example, let X = N(0,1). Let e = N(0, 1/32), e independent of X.


Let Y = 3 + 0.5 X + e.
Then (X,Y) are bivariate normal.
Y|X = (3+0.5X) + e which is normal since e is normal.

Find E(X), E(Y), var(X), var(Y), cov(X,Y), and r = cor(X,Y).


1.5 2.0 2.5 3.0 3.5 4.0 4.5
y

-3 -2 -1 0 1 2 3

x
4. Bivariate normal.

For example, let X = N(0,1). Let e = N(0, 0.22) and independent of X. Let Y = 3 + 0.5 X + e.

Find E(X), E(Y), E(Y|X), var(X), var(Y), cov(X,Y), and r = cor(X,Y).

E(X) = 0.
E(Y) = E(3 + 0.5X + e) = 3 + 0.5 E(X) + E(e) = 3.
Given X, E(Y|X) = E(3 + 0.5X + e | X) = 3 + 0.5 X. We will discuss this more later.
var(X) = 1.
var(Y) = var(3 + 0.5 X + e) = var(0.5X + e) = 0.52 var(X) + var(e) = 0.52 + 0.22 = 0.29.
cov(X,Y) = cov(X, 3 + 0.5X + e) = 0.5 var(X) + cov(X, e) = 0.5 + 0 = 0.5.
r = cov(X,Y)/(sd(X) sd(Y)) = 0.5 / (1 x .29) = 0.928.
In general, if (X,Y) are bivariate normal, can write Y = b1 + b2X + e, where E(e) = 0, and e
is ind. of X. Following the same logic, r = cov(X,Y)/(sxsy) = b2 var(X)/(sxsy) = b2 sx/sy,
so r = b s /s , and b = r s /s .
4. Bivariate normal.

For example, let X = N(0,1). Let e = N(0, 0.22) and independent of X. Let Y = 3 + 0.5 X + e.
In R,

4.5
4.0
x = rnorm(1000,mean=0,sd=1)

3.5
eps = rnorm(1000,mean=0,sd=.2)

3.0
y
y = 3 + .5*x+eps

2.5
plot(x,y)

2.0
cor(x,y) # 0.9282692.
1.5 -3 -2 -1 0 1 2 3

x
0.3
Density

0.2

95%
0.1

2.5% 2.5%
0.0

-4 -2 0 2 4

x
4. Bivariate normal.

If (X,Y) are bivariate normal with E(X) = 100, var(X) = 25, E(Y) = 200, var(Y) = 49, r = 0.8,
What is the distribution of Y given X = 105? What is P(Y > 213.83 | X = 105)?

Given X = 105, Y is normal. Write Y = b1 + b2X + e where e is normal with mean 0, ind. of X.
Recall b2 = r sy/sx = 0.8 x 7/5 = 1.12.
So Y = b1 + 1.12 X + e .
To get b1, note 200 = E(Y) = b1 + 1.12 E(X) + E(e) = b1 + 1.12 (100). So 200 = b1 + 112. b1 = 88.
So Y = 88 + 1.12 X + e , where e is normal with mean 0 and ind. of X.
What is var(e)?
49 = var(Y) = var(88 + 1.12 X + e) = 1.122 var(X) + var(e) + 2(1.12) cov(X,e)
= 1.122 (25) + var(e) + 0. So var(e) = 49 - 1.122 (25) = 17.64 and sd(e) = 17.64 = 4.2.
So Y = 88 + 1.12 X + e , where e is N(0, 4.22) and ind. of X.
Given X = 105, Y = 88 + 1.12(105) + e = 205.6 + e , so Y|X=105 ~ N(205.6, 4.22).
Now how many sds above the mean is 213.83? (213.83 205.6)/4.2 = 1.96,
so P(Y>213.83 | X=105) = P(normal is > 1.96 sds above its mean) = 2.5%.
5. Conditional expectation, E(Y | X), ch. 7.2.
Suppose X and Y are discrete.
Then E(Y | X=j) is defined as k k P(Y = k | X = j), just as youd think.
E(Y | X) is a random variable such that E(Y | X) = E(Y | X=j) whenever X = j.

For example, let X = the # of spades in your hand, and Y = the # of clubs in your hand.
a) Whats E(Y)? b) Whats E(Y|X)? c) Whats P(E(Y|X) = 1/3)?

a. E(Y) = 0P(Y=0) + 1P(Y=1) + 2P(Y=2)


=0+ 13x39/C(52,2) + 2 C(13,2)/C(52,2) = 0.5.

b. X is either 0, 1, or 2. If X = 0, then E(Y|X) = E(Y | X=0) and


E(Y | X=0) = 0 P(Y=0 | X=0) + 1 P(Y=1 | X=0) + 2 P(Y=2 | X = 0)
=0 + 13x26/C(39,2) + 2 C(13,2) / C(39,2) = 2/3.
E(Y | X=1) = 0 P(Y=0 | X=1) + 1 P(Y=1 | X=1) + 2 P(Y=2 | X = 1)
=0 + 13/39 + 2 (0) = 1/3.
E(Y | X=2) = 0 P(Y=0 | X=2) + 1 P(Y=1 | X=2) + 2 P(Y=2 | X = 2)
=0 + 1 (0) + 2(0) = 0.
So E(Y | X = 0) = 2/3, E(Y | X = 1) = 1/3, and E(Y | X = 2) = 0. Thats what E(Y|X) is
c. P(E(Y|X) = 1/3) is just P(X=1) = 13x39/C(52,2) ~ 38.24%.
6. Law of Large Numbers (LLN) and the Fundamental Theorem of Poker, ch 7.3.
David Sklansky, The Theory of Poker, 1987.
Every time you play a hand differently from the way you would have played it if
you could see all your opponents cards, they gain; and every time you play your
hand the same way you would have played it if you could see all their cards, they
lose. Conversely, every time opponents play their hands differently from the way
they would have if they could see all your cards, you gain; and every time they play
their hands the same way they would have played if they could see all your cards,
you lose.
Meaning?
LLN: If X1, X2 , etc. are iid with expected value and sd s, then ---> .
Any short term good or bad luck will ultimately become negligible to the sample mean.
However, this does not mean that good luck and bad luck will ultimately cancel out. See
p132.

You might also like