Random Variables Random Variables (Numeric)

Experiment Outcome Random Variable Range of

Random
§ Discrete random variable - can assume only Variable
Stock 50 Number of X = number of 0,1,2,, 50
a finite or limited set of values - i.e., the Xmas trees trees sold trees sold

Discrete
Inspect 600 Number Y = number 0,1,2,…,
number of automobiles sold in a year. items acceptable acceptable 600
Send out Number of Z = number of 0,1,2,…,
§ Continuous random variable - can assume 5,000 sales people people responding 5,000
letters responding
any one of an infinite set of values - i.e., Build an % completed R = % completed 0£R£100

Continuous
apartment after 4 after 4 months
temperature, product lifetime. building months
Test the Time bulb S = time bulb 0£S£80,000
lifetime of a lasts - up to burns
light bulb 80,000
1
(minutes) minutes

Probability Distributions
§Probability distribution – the set of all possible
values of a random variable and their associated
probabilities.

In a discrete probability distribution a probability

between 0 and 1 is assigned to each discrete variable.The
sum of the probabilities is equal to 1.

3 4
 Binomial Distribution
§ The binomial distribution is a probability
distribution with:
Øtrials that follow a Bernoulli process and have
two possible outcomes.
Øprobabilities that stay the same from one trial to
the next.
Øtrials that are statistically independent.
Øa positive integer number of trials.

5 6

Binomial Formulas Binomial Formulas (continued)

§ The binomial formula can be used to
§ For a binomial distribution, the expected value, or
determine the probability of r successes in n mean, is:
trials. The picture can't be displayed.
µ = np
§ The variance is:
= n! -
prqn r s 2 = np (1 - p )

Where,
r! (n - r)!
n = number of trials
r = number of successes
p = probability of success
q = probability of failure (1-p) 7 8
 Soda Selection:
Binomial Example Soda Selection Solution
§ Suppose 50% of your friends prefer diet soda to What is the probability that only one of your
regular soda. friends will select a diet soda? = .1563
What is the probability that three of your friends
§ You decide to practice your new binomial skills will select the diet soda? = .3125
while studying with five friends. These questions can be answered using the
§ You bring both diet and regular soda to your next binomial formula, where n = 5, r = 1 then 3, p = .5 and q = .5.
study session and offer one to each of your friends.
n!
P(r ) = p r q n-r
r!( n - r )!
Ø What is the probability that only one of your friends will
select a diet soda? 5!
Ø What is the probability that three of your friends will P (1) = 0.510.55 -1 = 0.1563
1!(5 - 1)!
select the diet soda? 5!
Ø What is the expected value, variance, and standard P (3) = 0.530.55 -3 = 0.3125
3!(5 - 3)!
deviation of your experiment?
9 10

Soda Selection Solution

Below is a graphical depiction of the answers.

0.40
P(r=3)
0.30

P(r=1)
P(r)

0.20

0.10

0.00
0 1 2 3 4 5
(r) Number of Successes

11 12
 The Binomial Distribution
The Binomial Distribution u In investments, the binomial distribution arises in connection with a
simplified model of the evolution of stock prices.

u Consider a stock whose closing price today is S.

• S1 , S 2 , … S n are the successive closing prices over the next n trading
days.
• Suppose on each day, the price either moves up to S *u or down to
S. *d
where S * is the previous day’s closing price, u and d are fixed
numbers such that d = 1/ u .
• The probability of an upward movement on each day is p,
independent of the stock price movements on the previous days.
13

The Binomial Distribution The Binomial Distribution

• Then the possible values of the stock price n days (i.e., the possible
values of Sn ) are

• The simplified model for stock prices just presented is known as the
binomial model for stock prices in the finance literature.
• A binomial tree for n=3 is displayed in Figure 5.2.
15 16
 The Poisson Distribution
§ The Poisson distribution is a discrete distribution
that is often used to describe arrival rates.
0.30
l=2
l x e -l
0.25
P( X ) =
0.20
X!
Expected value = l
Variance = l
0.15

0.10

0.05

0.00
1 2 3 4 5 6 7 8 9

17 18

The Poisson Distribution

19 20
 Normal Distribution

Probability density function - f(X)

5 5.05 5.1 5.15 5.2 5.25 5.3 5.35 5.4

-1 / 2 ( X - µ ) 2
1
f (X ) = s 2
e
s 2p
21 22

23 24
 Joint Distribution Bayesian Background
A bivariate distribution for a random vector X=(X1, X2)
with two components is an assignment of relative Prior Distribution Posterior Distribution
frequencies to the values and collections of values of X.

Question: If you are given the joint distribution as

follow: w w
f X ,Y ( x, y ) = xe - x ( y +1) , x, y > 0. Data Distribution Likelihood Function

Could we find the conditional densities? Say: f(x|y)

25
D w 26

27 28
29 30

31 32
 33 34

Joint Distribution Conditional densities

The joint density of X and Y is given by Then, we have
f X ,Y ( x, y ) = xe - x ( y +1) , x, y > 0.
f X ,Y ( x , y )
Determine the marginal densities. f X |Y = y ( x ) =
fY ( y )
Solution: =

and
f X ,Y ( x , y )
fY | X = x ( y ) =
f X ( x)
=

35 36
 Example
Law of Total Prob. An auto insurer is trying to develop a model for claim size
in an attempt to better price its products. On the basis of
historical data, it has determined that the claim size
¥ distribution for policyholders classified as good risks has
f X 1 ( x1 ) = ò -¥
f X 1 | X 2 =t ( x1 ) f X 2 (t ) dt
density
¥
f X 2 ( x2 ) = ò f X 2 | X 1 = s ( x2 ) f X 1 ( s ) ds f X ( x ) = 2e -2 x , x >0
-¥

and the claim size distribution for policyholders classified

as bad risks has density
1 -x /3
f X ( x) = e , x > 0,
3
where claims are measured in thousands of dollars. There is
a 30% chance that a given policyholder is a bad risk. What
is the probability that this policyholder’s claim exceeds
37 38
$1000?

Let X be the claim size and let I be an indicator for By the law of total probability:
being a bad risk. Then, from the given information:
f X ( x) =
ì1 with probability 0.3
I =í
î0 with probability 0.7

And further,
and
f X |I =0 ( x ) = 2e -2 x , x > 0
¥
f X |I =1 ( x ) =
1 -x / 3
e , x > 0.
P( X > 1) = ò f X ( x )dx
1
3

39 40
 Example 2
An insurance company has just sold a group health insurance plan to a Let N be the no. of hospitalization claims in the next month. Then,
new employer. Experience with similar employers suggests that the
number of hospitalization claims per month can be modeled using the ln e - l
p N |L = l ( n ) = , n = 0,1,2,...
mass function n!
ln e - l and
p N ( n) = , n = 0,1,2,...
n! f L (l ) = e - l , l > 0.

where λ is a parameter that describes the group’s expected utilization. By the law of total prob.,
The insurer is uncertain what the true value of λ for this group is (i.e.,
the insurer is uncertain about how many claims to expect) and decides
to model this uncertainty using the density function
f L (l ) = e - l , l > 0.

Determine the probability that the group has two or more 41 42

hospitalization claims in the next month.

Bayes’ Theorem
Hence the desired prob. is Further, suppose that there is one hospitalization in the first month.
How does this information alter the insurer’s belief about the true
value of λ?
At the end of the first month, the insurer’s belief about the the
parameter λ is captured by the conditional density.
Using Bayes’ theorem, we know that
f N |L = l (1) f L (l )
f L| N =1 (l ) =
f N (1)
There is a 25% chance that the group will have more than 1
p N |L = l (1) f L (l )
hospitalization claims next month. =
p N (1)
(le - l )(e - l )
= .
p N (1)
43 44
 Bayes’ Theorem Bayes’ Theorem
By the law of total prob., If insurer’s belief about the parameter λ exceeding two. We have,
¥
pN (1) = ò p l (1) f (l )d l
0 N |L = L
p ( L > 2) = ò 2
¥
e - l dl =e - 2 .

= ò ( le l ) e l d l
¥
- -
0
At the end of the month, insurer knew that there is one
= hospitalization claimed, the insurer believer that the chance of λ
exceeding two is:
¥
Hence, p ( L > 2 | N = 1) = ò 4l e -2 l d l =
2

f L| N =1 (l ) =
45 46

47 48
49 50

51 52