Quantitative Analysis

CFA Level I

By Dr. Jatin Thukral

 Outline
• Probability Concepts
• Common Probability Distributions
• Sampling and Estimation
• Hypothesis Testing
 Probability Concepts
• A random variable is an uncertain quantity
• An outcome is an observed value of a random
variable
• An event is a single outcome or a set of
outcomes
• Mutually exclusive and exhaustive events are
those that include all possible outcomes
 Defining Properties of Probability
• The probability of occurrence of any event Ei is
between 0 and 1
0 ≤ P(Ei) ≤ 1
• If a set of event E1 ,E2 , E3 , …. , EN is mutually
exclusive and exhaustive, then
∑P(E ) = 1
i
 Types of Probabilities

• Empirical probability: based on historical data

• A priori probability: based on reasoning

• Subjective probability: based on personal

judgment
 Probability in terms of Odds

• Alternate way of expressing probabilities

• The odds that an event with probability 0.2

will occur are 0.2/(1-0.2) = 0.25
Unconditional and Conditional Probabilities

• Unconditional Probability: The marginal

probability that an event occurs regardless of
any past or future event

• Conditional Probability: The probability of any

event given that another event(s) has
occurred or will occur
An example illustrating unconditional, conditional and joint
probabilities
 Rules of Probability
• Multiplication Rule:
Joint Probability P(AB) = P(A|B) P(B)
• Addition Rule:
P( A or B) = P(A) + P(B) – P(AB)
• Total Probability Rule:
P(A) = P(A|B1) P(B1) + P(A|B2) P(B2)
+ ………. + P(A|BN) P(BN)
where B1, B2 ,…, BN is a mutually exclusive and
exhaustive set of outcomes
 Applications of Probability Rules
• Multiplication rule is used to compute joint
probability
• Addition rule is used to determine probability
that at least one of two event occurs
• Joint probability of any number of
independent events is calculated as
P(ABCD)=P(A) P(B) P(C) P(D)
 Independent Events
• Independent Events: Occurrence of one event
has no influence on occurrence of other events
P(A|B) = P(A)
 Expected Value

Expected value is the weighted average of

the possible outcomes of a random variable,
where the weights are the probabilities that
the outcomes will occur
Diagramming an investment problem using
a tree diagram
 Covariance

• Covariance is a measure of how two assets

move together
• Variance is covariance of an asset with itself
• Covariance can range from negative infinity to
positive infinity
 Correlation Coefficient

• Correlation is covariance divided by product of

standard deviations
• Measures the strength of linear relationship
between two random variables
• Has no units and ranges from -1 to +1
• If Corr(Ri,Rj)=1, the random variables have perfect
positive correlation
 Portfolio Expected Value and Variance

• Portfolio Expected Value

• Portfolio Variance
 Bayes’ Formula
• Bayes’ formula is used to update a given set of prior
probabilities for a given event in response to the arrival of
new information
 Labeling
• There are n items and • The total number of
each can receive one of ways that the labels can
k different labels be assigned is
• The number that
receive label 1 is n1, that
receive label 2 is n2 and
so on, such that
 Combination Formula
• A special case of labeling where k=2
• n1=r and n2=n-r
• Choosing r items from a set of n items
• nCr is the number of possible ways of selecting r
items from n items when order of selection is
not important
 Permutation Formula
• nPr is the number of possible ways to select r
items from a set of n items when the order of
selection is important
 Outline
• Probability Concepts
• Common Probability Distributions
• Sampling and Estimation
• Hypothesis Testing
 Probability Distribution
• A probability distribution describes the
probabilities of all the possible outcomes for a
random variable
• A discrete random variable is one for which
number of possible outcomes can be counted
• A continuous random variable has infinite
possible outcomes
 Probability Function
• A probability function p(x) specifies the
probability that a random variable is equal to
a specific value x

• The two key properties are

0 ≤ p(x) ≤ 1
∑ p(x) = 1
 Probability Density Function (pdf) and
Cumulative Distribution Function (cdf)

• A pdf is a function f(x) that generates the

probability that outcomes of a continuous
distribution lie within a particular range

• A cdf defines the probability that a random

variable X takes on a value less than a specific
value x
Pdf and Cdf of Normal Distribution
 Discrete Uniform Random Variable and a
Binomial Random Variable
• A discrete uniform random variable:
probabilities for all possible outcomes are
equal
• A binomial random variable is the number of
“successes” in a given number of trials when
outcome can be either “success” or “failure”
 Binomial Random Variable
• The probability of success p is constant for
each trial and the trials are independent
• For number of trials = 1, the binomial random
variable is called Bernoulli random variable
• The expected value of a Binomial variable is
Stock Price Movements as Binomial Tree
 Continuous Uniform Distribution
• Outcomes can only occur between some
lower limit a and some upper limit b
• P(X < a) = P(X > b) = 0
• P(x1 ≤ X ≤ x2) = (x2 – x1) / (b – a)
 Normal Distribution
• Completely described by mean µ and variance
σ2, stated as X ~ N(µ,σ2)
• Skewness = 0, that is P(X ≤ µ) = P(µ ≤ X) = 0.5
• Kurtosis = 3
• A linear combination of normally distributed
random variable is also normally distributed
• Probabilities of outcomes further above and
below from mean get smaller but never zero
Normal Distribution
 Multivariate Distributions
• Specifies the probabilities associated with a
group of random variables
• Meaningful only if behavior of one random
variable depends on behavior of others
• Multivariate normal distribution is completely
defined by three set of parameters
 Confidence Interval
• A confidence interval is a range of values
around the expected outcome within which
we expect the actual outcome to be some
specified percentage of the time
 Standard Normal Distribution
• A standard normal distribution is a normal
distribution that has been standardized so that
it has mean zero and standard deviation 1
• Standardization is the process of converting an
observed value for a random variable to its z-
value
 Using the z-table
• Consider an EPS distributed with µ=6,σ =2.
What is probability that EPS ≥ 9.70?
• The z-value of EPS = 9.70 is

• From z-table, F(1.85)=0.9678

• F(EPS > 9.70) = 1-0.9678 = 0.0322
Shortfall Risk and Safety-First Ratio
• Shortfall risk is probability that return will fall
below a target value over a given time period
• Roy’s safety-first criterion states that optimal
portfolio minimizes the probability that return
falls below a minimum threshold level, i.e.,

• If returns are normally distributed, then

 Lognormal Distributions
• The lognormal distribution is generated by
the function ex where x is normally distributed
 Monte Carlo Simulations
• Technique based on repeated generation of
one or more factors that affect security values
• A probability distribution is assigned to each
factor
• Each set of randomly generated factors is used
with a pricing model to value the security
• Limitations: fairly complex; results no better
than the assumptions on factor distributions
 Historical Simulation
• Based on actual changes in risk factors over
some time period
• Advantage of using the actual distribution of
risk factors and not estimations
• Limitation compared to Monte Carlo
simulation is that can’t answer “What if”
questions
 Outline

• Probability Concepts
• Common Probability Distributions
• Sampling and Estimation
• Hypothesis Testing
 Sampling
• Simple random sampling: selecting a sample such
that each item has same chance of being selected
• Systematic sampling: selecting every mth item
• Sampling error: difference between sample statistic
(mean, variance) and its population parameter
• Sampling distribution of a sample statistic is a
probability distribution of all possible sample statistics
computed from a set of equal-sized samples
 Stratified Random Sampling
• Population is separated into smaller groups
based on one or more distinguishing
characteristics
• From each group, or stratum, a random
sample is taken and results are pooled
• Size of sample from each stratum depends on
size of stratum relative to population
• Often used in bond indexing
 Time-Series and Cross-Sectional Data

• Times-series data: observations taken over a

period of time at specific and equally spaced
time intervals
• Cross-sectional data: observations taken at a
single point in time
 Central Limit Theorem
• For simple random samples of size n from a
population with mean µ and variance σ2, the
sampling distribution of sample mean
approaches N(µ,σ2/ n)
• Extremely useful: Normal distribution is easy to
apply to hypothesis testing and construct
confidence intervals
• Inference can be made irrespective of
population distribution if n>30
 Standard Error of Sample Mean
• It is the standard deviation of the distribution
of the sample means
Point Estimates and Confidence Interval
• Point estimates are single values used to estimate
population parameters. For ex.

• Confidence interval estimates result in a range

of values within which the actual parameter will
lie with probability 1-α. Here, α is called the
level of significance and 1-α is called degree of
confidence
 Construction of Confidence Intervals
point estimate ± (reliability factor X standard error)
Where
Point estimate = vale of a sample statistic
Reliability factor = a number depending on the
sampling distribution of estimation
and probability that estimate falls within
confidence interval
Standard error = standard error of the point estimate
Desirable Properties of an Estimate
• Unbiased: Expected value of the estimator is
equal to the parameter you are trying to
estimate
• Efficient: Variance of sampling distribution is
smaller than all the other unbiased estimators
• Consistent: Accuracy of parameter estimate
increases as the sample size increases
 Student’s t-distribution
• Bell-shaped distribution, symmetric about mean
• Appropriate distribution to use when
constructing confidence intervals based on small
samples (n<30) from populations with unknown
variance and a normal (or approx.) distribution
• Also appropriate when variance is unknown and
sample size is large enough that CLT will ensure
that sampling distribution is approx. normal
 Properties of Student’s t-distribution

• Symmetrical about mean

• Defined by a single parameter, the degrees of
freedom, equal to n – 1
• More probability in tails than normal
distribution
• As degrees of freedom get larger it
approaches standard normal distribution
• Fatter tails => difficult to reject null hypothesis
Distributions for Different Degrees of
Freedom
 Calculating Confidence Interval
• Population is normal distributed with known
variance
 Calculating Confidence Interval

• Population is normal distributed with

unknown variance
 Constructing Confidence Intervals
• Unlike standard normal distribution, the
reliability factor for t-distribution depends on
sample size, so we have to look for reliability
factors from table
• Owing to fatter tails of the t-distribution,
confidence intervals using t-distribution are
wider/more conservative
Selecting the Appropriate Test Statistic
 Biases in Estimation Based on Samples
• Data-Mining Bias: Repetitive search for data patterns
results in data-mining
• Sample Selection Bias: Some data is systematically
excluded because of lack of availability
• Survivorship Bias: Most common bias, only surviving
data is considered
• Look-Ahead Bias: A relationship is tested based on
data not available on test date
• Time-Period Bias: Period over which data is gathered is
too short or too long
 Outline

• Probability Concepts
• Common Probability Distributions
• Sampling and Estimation
• Hypothesis Testing
Hypothesis Testing Procedure
Null Hypothesis and Alternative Hypothesis

• The null hypothesis H0 is the hypothesis that

the researcher wants to reject. For ex,

• The alternate hypothesis Ha is what is

concluded if there is sufficient evidence to
reject the null hypothesis
 Two-Tailed Tests of Hypotheses
• A two-tailed test for population mean is

• The general decision rule for a two-tailed test is

 Two-Tailed Test Using Standard Normal(z)
Distribution
• The decision rule for a two-tailed z-test at α=0.05 is
 Example of Two-Tailed Test
• Data gathered on daily returns for 250 days;
Mean daily return=0.1%, Sample Std. Dev.
=0.25%. Belief: Mean Daily Return ≠ 0
• Null Hypothesis:
• Decision Rule:
• Standard error = 0.0025/√250
• Test Statistic = (0.001 – 0)/(0.0025/√250) = 6.33
• Since 6.33 > 1.96, reject the null hypothesis
 One-Tailed Hypothesis Test
• The null and alternative hypotheses are either

• Using Standard Normal(z) Distribution

 Test Statistic

• Standard Error when population σ is known:

• Standard Error when population σ is not known:

 Critical Value and Decision Rule
• A test statistic is a random variable that may
follow one of several distributions: the t-
distribution, the z-distribution, the chi-square
distribution and the F-distribution
• The critical value for the appropriate test
statistic depends on its distribution
• The decision rule: if the test statistic is
(greater, less than) X, reject the null
 Type I and Type II Errors
• Type I error: Rejection of H0 when it is actually true
• Type II error: Failure to reject H0 when it is false
• The significance level is the probability of making a
Type I error and is designated by α
• Power of a test is the probability of correctly
rejecting H0 when it is false:
• Therefore, Power of a test = 1 – P(Type II error)
Type I and Type II Errors
 Relation Between Confidence Intervals and
Hypothesis Tests
• A confidence interval is determined as

• The above expression can be rewritten as

 p-value and Hypothesis Testing
• The p-value is the smallest level of significance
for which H0 can be rejected assuming it is
true
• For example,
Signif. level =95%,
Test statistic =2.3
 Population mean: when to use t-test?
• Use t-test when the population variance is
unknown and either n≥30 or n<30 but
distribution is approx. normal
• A t-statistic with n-1 degrees of freedom is
computed as
 Population mean: when to use z-test?
• Use z-test when population is normally
distributed and variance is known
• A z-statistic is computed as
 Population mean: when to use z-test?
• Also, use z-test when sample size is large and
variance is unknown
• Corresponding z-statistic is computed as
Critical z-Values
 Equality Test of Two Population Means
• Three possible hypotheses structures

• Two possible t-statistics

• Unknown variances but assumed equal
• Unknown variances assumed unequal
Unknown Variances Assumed Equal
Unknown Variances Assumed Uequal
 Paired Comparisons Test
• Test whether the means of the difference
between observations for two samples are
different
• Two-sided Test

• One-sided Test
t-statistic for Paired Comparisons Test
 Test Variance of Normally-Distributed
Population
• Chi-square test
– Two-tailed test

– One-tailed test
Decision Rule for a Two-Tailed Chi-
Square Test
 Chi-Square Test Statistic
• The chi-square test statistic with n-1
degrees of freedom is computed as
 Equality of Variance of Two Normally
Distributed Populations
• F-distributed test statistic
– Two-tailed test

– One-tailed test

• F-statistic computed as
 F-test
• Always put larger variance in numerator
• n1 – 1 and n2 – 1 are degrees of freedom used
to identify critical value from the F-table
 Parametric and Nonparametric Tests

• Parametric Tests: Rely on assumptions

regarding the distribution of the population
and are specific to population parameters
• Nonparametric Tests: Either do not consider a
particular population parameter or have few
assumptions about the population that is
sampled