Académique Documents
Professionnel Documents
Culture Documents
com
Quantitative Analysis - I
www.edupr i st i ne.com
Agenda
Probability
Counting Principle(1hr)
Probability Concepts, Bayes Theorem(1hr)
Moments
Probability distributions
Neev Knowl edge Management Pr i st i ne
2
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Counting Principle
Number of ways of selecting r objects out of n objects
n
C
r
n!/(r!)*(n-r)!
Number of ways of giving r objects to n people, such that repetition is allowed
N
r
3
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
4
Example-Counting Principle
In how many ways 3 stocks can be chosen out of 10 stocks in a portfolio?(Combination)
Choosing 3 out of 10 stocks is basically the number of combinations of 3 objects out of 10
Therefore, the number of ways are
10
C
3
= =120
In how many ways 3 stocks can be sold, if the sold stock is bought back in the portfolio before the
next stock is sold?
First stock can be sold in 10 ways
Second can be sold in again 10 ways
Third stock can again be sold in 10 ways
Therefore total number of ways become =10
3
=1000
)! 3 10 !*( 3
! 10
) , (
) , ( =
31
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
32
Some Properties of Variance
Variance of a constant = 0
Covariance between same variables is also their
variance
For independent or uncorrelated variables,
covariance or correlation = 0
) ( ) (
2
X Var a b aX Var = +
) , ( 2 ) ( ) ( ) (
2 2
Y X abCov Y Var b X Var a bY aX Var + + = +
= =
=
n
i
i
n
i
i
X Var X Var
1 1
) ( ) (
= = =
=
n
i
n
j
j i
n
i
i
X X Cov X Var
1 1 1
) , ( ) (
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
33
Skewness
Skewness describes departures from symmetry
Skewness can be negative or positive.
Negative skewness indicates that the distribution has a
long left tail, which indicates a high probability of observing
large negative values.
If this represents the distribution of profits and losses for
a portfolio, this is a dangerous situation.
3
1
3
) (
=
=
n
i
i
k
x
S
Normal Distribution
Positively Skewed Distribution
Negatively Skewed Distribution
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
34
Kurtosis
Kurtosis describes the degree of flatness of a distribution, or width of its tails.
Because of the fourth power, large observations in the tail will have a large weight and hence create
large kurtosis. Such a distribution is called leptokurtic, or fat tailed.
A kurtosis of 3 is considered
average.
High kurtosis indicates a
higher probability of extreme
movements.
4
1
4
) (
=
=
n
i
i
x
K
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
-4 -3 -2 -1 0 1 2 3 4
Platykurtic
K<3
Mesokurtic
K=3
Leptokurtic
K>3
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
35
Example
Assuming that the distribution of ABC stock returns is a population, what is the population variance
and standard deviation?
A. 5.0
B. 6.8
C. 45.22
D. 80.2
ABC Annual stock prices
2001 2002 2003 2004 2005 2006
12% 5% -7% 11% 2% 11%
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
B
The population variance is given by taking the mean of all squared deviations from the mean.
2
= [(12-5.67)
2
+ (5-5.67)
2
+ (-7-5.67)
2
+ (11-5.67)
2
+ (2-5.67)
2
+ (11-5.67)
2
] / 6
= 45.22 (%
2
)
The standard deviation is the square root of the variance:
% 72 . 6 % 22 . 45 = =
36
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
The random variables X and Y have variances of 2 and 3 respectively, and covariance of 0.5.
The variance of 2X + 3Y is:
A. 13
B. 29
C. 35
D. 41
Example
37
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
D
Var(X + Y) = Var(X) + Var(Y) +2*Cov(x,y)
Var(X - Y) = Var(X) + Var(Y) -2*Cov(x,y)
Var(cX) = c^2 * Var(X)
Cov (ax,by) = abCov(x,y)
So, Var(2X + 3Y) = 2
2
Var(X) +3
2
Var(Y) +2*2*3*Cov(x,y)
Var(2X + 3Y) = 4*2 + 9*3 + 12*0.5 = 41
Solution
38
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
39
Question
You are given the following information about the returns of stock P and stock Q:
Variance of return of stock P=100.0
Variance of return of stock Q=225.0
Covariance between the return of stock P and the return of stock Q=53.2
At the end of 1999, you are holding USD 4 million in stock P. You are considering a strategy of
shifting USD 1 million into stock Q and keeping USD 3 million in stock P. What percentage of risk,
as measured by standard deviation of return, can be reduced by this strategy?
A. 0.50%
B. 5.00%
C. 7.40%
D. 9.70%
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
40
Solution
B
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
41
FRM Exam 2009
Which type of distribution produces the lowest probability for a variable to exceed a specified extreme
value X which is greater than the mean assuming the distributions all have the same mean and
variance?
A. A leptokurtic distribution with a kurtosis of 4.
B. A leptokurtic distribution with a kurtosis of 8.
C. A normal distribution.
D. A platykurtic distribution
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
D
By definition, a platykurtic distribution has thinner tails than both the normal distribution and any leptokurtic
distribution. Therefore, for an extreme value X, the lowest probability of exceeding it will be found in the
distribution with the thinner tails.
A. Incorrect. A leptokurtic distribution has fatter tails than the normal distribution. The kurtosis indicates
the level of fatness in the tails, the higher the kurtosis, the fatter the tails. Therefore, the probability of
exceeding a specified extreme value will be higher .
B. Incorrect. Since answer A. has a lower kurtosis, a distribution with a kurtosis of 8 will necessarily
produce a larger probability in the tails.
C. Incorrect. By definition, a normal distribution has thinner tails than a leptokurtic distribution and larger
tails than a platykurtic distribution.
42
www.edupr i st i ne.com
Agenda
Probability
Moments
Probability distributions
Discrete Probability Distributions (1hr)
Continuous Distribution (1hr)
Neev Knowl edge Management Pr i st i ne
43
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Probability Distribution
A Random Variable is a function, which assigns unique numerical values to all possible outcomes of
a random experiment under fixed conditions. A random variable is not a variable but rather a function
that maps events to numbers
Probability distribution describes the values and probabilities that a random event can take place. The
values must cover all of the possible outcomes of the event, while the total probabilities must sum to
exactly 1, or 100%
Example
Suppose you flip a coin twice.
There are four possible outcomes: HH, HT, TH, and TT.
Let the variable X represent the number of Heads that result from this experiment
It can take on the values 0, 1, or 2.
X is a random variable (its value is determined by the outcome of a statistical experiment)
A probability distribution is a table or an relation that links each outcome of a statistical experiment with
its probability of occurrence
Number of heads (X) Probability P(X=x)
0 0.25
1 0.50
2 0.25
44
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
45
Continuous & Discrete Probability Distributions
If a variable can take on any value between two specified values, it is called a continuous variable
otherwise, it is called a discrete variable
If a random variable is a discrete variable, its probability distribution is called a discrete probability
For example, tossing of a coin & noting the number of heads (random variable) can take a discrete
value
Binomial probability distribution, Poisson probability distribution
If a random variable is a continuous variable, its probability distribution is called a continuous
probability distribution
The probability that a continuous random variable will assume a particular value is zero
A continuous probability distribution cannot be expressed in tabular form.
An equation or formula is used to describe a continuous probability distribution (called a probability
density function or density function or PDF)
The area bounded by the curve of the density function and the x-axis is equal to 1, when computed over
the domain of the variable
Normal probability distribution, Student's t distribution are examples of continuous probability
distributions
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
46
Probability distribution
Cumulative Probability is a rule or equation which describes the sum of all the probabilities till that
observation.
Take the previous example of flipping of coin twice. The following table gives the probability of
occurrence of heads and the cumulative probability as well.
The point to note here is that the cumulative probability of the first event is equal to the probability of
that event.
The cumulative probability of the last event is always 1
Number of heads (X) Probability P(X=x) Cumulative Probability: P(X < x)
0 0.25 0.25
1 0.50 0.75
2 0.25 1
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
47
Bernoulli Distribution
Assumptions
The outcome of an experiment can either be success (i.e., 1) and failure (i.e., 0).
Pr(X=1) = p, Pr(X=0) = 1-p, or
The expected value E[X] of the event is equal to the probability of success(p)
E[X] = p
The variance of the event is the product of the probability of success and probability of failure:
Var(X) = p(1-p)
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
48
Binomial Distribution
Assumptions:
There are n trials.
Each trial has two possible outcomes, success or failure.
The probability of success p is the same for each trial.
Each trial is independent.
If we take n Bernoulli trials, and say out of those n trials we have total number of x successes,
then the probability of such an event can be given as:
The expected number of successes E[X] = n * p
The variance of number of successes Var (X) = n * p * (1-p)
) (
) 1 ( * * ) (
x n x
x
n
p p C x X P
= =
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
49
Example
There are 10 bonds in a credit default swap basket; the probability of default for each of the bonds
is 5%. The probability of any one of the bond defaulting is completely independent of what happens
to the other bonds in the basket. What is the probability exactly one bond default?
A. 5%
B. 50%
C. 32%
D. 3%
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
C
one particular bond defaults and other nine do not with the probability 0.05* (0.95)^9 can happen in 10
different ways.
= 10 * 0.05^1* (0.95)^9 = 32%
50
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
51
Company ABC was incorporated on January 1, 2004. it has expected annual default rate of 10%.
Assuming a constant quarterly default rate, what is the probability that company ABC will not have
defaulted by April 1, 2004?
A. 2.4%
B. 2.5%
C. 97.4%
D. 97.5%
Question
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
C
Let the probability of not defaulting in 1 quarter is (nd). Then the probability of not defaulting for a full
year is (nd)
4
. This implies that the probability of defaulting within 1 year time is {1-(nd)
4
}, which is
given as 10%.
1-(nd)
4
=0.1 which implies
(nd)=0.9
1/4
= 97.4%
52
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
53
Question
A corporate bond will mature in 3 years. The marginal probability of default in year one is 0.03%.
The marginal probability of default in year 2 is 0.04%. The marginal probability of default in year 3
is 0.06%. What is the cumulative probability that default will occur during the 3 year period?
A. 0.1247%
B. 0.1276%
C. 0.1299%
D. 0.1355%
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
54
Solution
C
The cumulative probability of default= 1-{Product of marginal probabilities of not defaulting}
=1-{(1-0.0003)*(1-.0004)*(1-0.0006)}
=0.001299
Therefore the cumulative probability of default is 0.1299%
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Poisson Distribution
Assumptions:
The probability of observing a single event over a small interval is approximately proportional to the size
of that interval.
The probability of an event within a certain interval does not change over different intervals.
The probability of an event in one interval is independent of the probability of an event in any other
interval which is not overlapping.
Poisson distribution is a special case of Binomial distribution when the probability of success (p)
becomes very small and the number of events (n) becomes very large in such a way that the
product of both gives a constant ().
Fix the expectation =n * p
Number of trials n
A Binomial distribution will become a Poisson distribution
E[X] = , Var(X) =
>
= =
otherwise 0
0
!
) (
x e
x
x X P
x
55
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
56
Plots of Poisson Distribution
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0 2 4 6 8 10 12 14 16 18 20 22 24
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
0 2 4 6 8 10 12 14 16 18 20 22 24 26 28
0
0.02
0.04
0.06
0.08
0.1
0.12
0 2 4 6 8 10 12 14 16 18 20 22 24
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54
=1
=5
=25 =15
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
57
Question
When can you use the Normal distribution to approximate the Poisson distribution, assuming you
have "n" independent trials each with a probability of success of "p"?
A. When the mean of the Poisson distribution is very small.
B. When the variance of the Poisson distribution is very small.
C. When the number of observations is very large and the success rate is close to 1.
D. When the number of observations is very large and the success rate is close to 0.
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
C
The Normal distribution can approximate the distribution of a Poisson random variable with a large lambda
parameter (). This will be the case when both the number observations (n) is very large and the success
rate (p) is close to 1 since = n*p.
INCORRECT: A, The mean of a Poisson distribution must be large to allow approximation with a Normal
distribution.
INCORRECT: B, The variance of a Poisson distribution must be large to allow approximation with a
Normal distribution.
INCORRECT: D, The Normal distribution can approximate the distribution of a Poisson random variable
with a large lambda parameter (). But since = n*p, where n is the number observations and p is the
success rate, will not be large if p is close to 0.
58
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
59
Example
The number of false fire alarms in a suburb of Houston averages 2.1 per day. What is the
(approximate) probability that there would be 4 false alarms on 1 day?
A. 1
B. 0.1
C. 0.5
D. 0
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
60
Solution
B
Use Poisson distribution
P(X = x) = [x *e-]/ !
Is there any other intuitive way as well???
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
61
Uniform distribution
The simplest continuous distribution function is the uniform distribution. This is defined over a
range of values for x, a x b.
The density function is,
f (x) =1 /(b a) , a x b
Its mean and variance are given by
E(X) =(a + b)/ 2
V(X) = (b a)
2
/12
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
62
Question
Assume we use continuous uniform distribution U(0,10) to generate a series of random numbers.
Which of the following statements is Correct?
A. The number 5 is likely to be observed much more often than any other number.
B. Numbers between 4 and 6 are more likely to occur than the number between 6 and 10, because the first
interval is closer to the center of the distribution.
C. Numbers between 1 and 3 are as likely as the number between 4 and 6.
D. Numbers between 1 and 3 are less likely than the number between 4 and 6, due to skewness of the
distribution.
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
63
Solution
C
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
64
Standard Normal Distribution
The standard normal distribution has mean = 0 and standard deviation sigma=1
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Normal (Gaussian) Distribution
The normal distribution is defined by first two moments, mean (u) and variance (o
2
)
The probability density function P(x) of normally distributed variable is given by:
The probability of the value lying between a and b is given by:
The expected value of a normally distributed variable: E[X]= u,
The variance of normally distributed variable: Var(X)= o
2
If two variables are individually normally distributed, then the linear combination of the both is
also normally distributed.
Lets take an example of two variable X
1
and X
2
which are normally distributed such that:
X
1
~N(u
1
,o
1
) and X
2
~N(u
2
,o
2
)
Then X= a.X
1
+ b.X
2
is also normally distributed.
)
`
=
2
2
2
2
) (
exp
2
1
) (
x
x P
= s s
b
a
dx x P b X a P ). ( ) (
The skewness of normal distribution is = 0 and the kurtosis is =3
65
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
66
Question
Suppose the standard deviation of a normal population is known to be 10 and the mean is
hypothesized to be 8. Suppose a sample size of 100 is considered. What is the range of sample
means that allows the hypothesis to be accepted at a level of significance of 0.05?
A. Between -11.60 and 27.60
B. Between 6.04 and 9.96
C. Between 6.355 and 9.645
D. Between -8.45 and 24.45
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
B
To accept the hypothesis at a 0.05 significance level, the test statistic Z must fall between -1.96 and 1.96
Z = (X - 8) / (10 / Sqrt(100)) and -1.96 <=z<=1.96
which implies that the sample mean X must be between 6.04 and 9.96.
67
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
68
Example: FRM EXAM 2002
Consider a stock with an initial price of $100. Its price one year from now is given by S = 100exp(r),
where the rate of return r is normally distributed with a mean of 0.1 and a standard deviation of 0.2.
With 95% confidence, after rounding, S will be between
A. $67.57 and $147.99
B. $70.80 and $149.20
C. $74.68 and $163.56
D. $102.18 and $119.53
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
C.
Note that this is a two-tailed confidence band, so that = 1.96.
We find the extreme values from $100exp( ).
The lower limit is then V1 = $100 * exp(0.10 1.96 * 0.2) = $100 * exp(0.292) = $74.68. The upper limit
is
V2 = $100 * exp(0.10 + 1.96 0.2) = $100 * exp(0.492) = $163.56.
69
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
70
Question
Let X be a uniformly distributed random variable between minus one and one so that the standard
deviation of X is 0.577. What percentage of the distributions will be less than 1.96 standard
deviations above the mean:
A. 100%
B. 97.5%
C. 95%
D. Insufficient information provided.
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
A
The answer requires understanding of distributions and standard deviation. The key is that every
distribution has a standard deviation. However the number of standard deviations associated with
different probabilities are different for each distribution. In this case 1.96 standard deviation represents
a move of 1.12 or less. As the total distribution is defined as falling between minus one and one the
correct answer is A.
71
www.edupr i st i ne.com
Question
For the standard normal distribution, calculate the value of P(-1.87 Z 1.23) or P(|Z| 1.6)?
A. 0.5683
B. 0.8794
C. 0.7831
D. 0.9145
Neev Knowl edge Management Pr i st i ne
72
www.edupr i st i ne.com
Solution
D
In the diagram given below, the area representing the region P(-1.87 Z 1.23) or P(|Z| 1.6) is
shown below. The area will be from Z = -1.87 to Z = 1.6 and common area is from Z = 1.6 to Z = 1.23.
P(-1.87 Z 1.23) or P(|Z| 1.6) = P(Z 1.87) + P(Z 1.6) = 0.4693 + 0.4452 = 0.9145
Hence option D is correct.
Neev Knowl edge Management Pr i st i ne
73
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
74
Agenda
Appendix
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Appendix: Lognormal Distribution
A random variable X is said to have a lognormal distribution if its logarithm Y = ln(X) is normally
distributed.
The lognormal density function has the following expression:
)
`
=
2
2
2
2
) ) (ln(
exp
2
1
P(x)
x
x
75
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
76
Appendix: The Chi-square Distribution
The chi-square distribution is a family of distributions, depending on degrees of freedom:
d.f. = n - 1
0 4 8 12 16 20 24 28
d.f. = 15
2
0 4 8 12 16 20 24 28
d.f. = 5
2
0 4 8 12 16 20 24 28
d.f. = 1
2
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Assume you have empirical data showing historical returns (v) for a given financial variable
(e.g.: Forex rate), how could you perform a quick test of the validity of the power law
Prob(v > x) = K * x
-a
where x is large, as a good model of the tail of the distribution?
A. Plot the probability of v exceeding x standard deviations against x
B. Plot the probability of v exceeding x standard deviations against Log of x
C. Plot the Log of the probability of v exceeding x standard deviations against x
D. Plot the Log of the probability of v exceeding x standard deviations against the Log of x
Appendix: Example FRM EXAM
77
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
D
The mathematical relationship in the question can be rewritten (by taking the logs on both sides):
Log(Prob(v > x)) = Log(K) - aLog(x), i.e. the plot of the Log of the probability of v exceeding x standard
deviations against the log of x should be a straight (decreasing) line if the relationship strictly holds.
The intercept is an estimate of Log of K and the slope of the line yields the parameter a.
78
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
79
Appendix: Sum Rule & Bayes Theorem
Jack has 3 white balls and 2 red balls in his box while his friend Andrew has 4 white and 5 red balls.
Andrew took 1 white ball from jack and gave him 1 red ball in compensation. Now calculate the
probability of picking a red ball from Andrews box.
After the exchange, Tom stole a ball from one of the boxes and found that its white. If you have to tell
who lost his white ball, what should be your say?
www.edupr i st i ne.com Neev Knowl edge Management Pr i st i ne
Solution
Initially Jack has 3 white balls and 2 red balls in his box while his friend Andrew has 4 white
and 5 red balls.
After the exchange Jack has 2 white and 3 red balls, and Andrew has 5 white and 4 red balls.
Therefore the probability of picking a red ball from Andrews box is:
P(R
Andrew
)=4/(5+4)=4/9
Now Tom stole a white ball from one of the two boxes. To make a calculated guess about
who lost 1 white ball, we need to calculate the conditional probabilities.
P(Jacks box/If the balls is White)= Probability of white balls in Jacks box/(Probability of white
ball in Jacks box +Probability of white ball in Andrews box)
10
P
3
= 720
Solution