CFA LEVEL 1

QUANTITATIVE METHOD

Lecturer: Tran Trong Kien, CFA

Email: Trankien88@[Link]

1
 Quantitative
8-12%

READING 1: RATE AND RETURNS

READING 2: TIME VALUE OF MONEY IN FINANCE
READING 3: STATISTICAL MEASURES OF ASSET RETURNS
READING 4: PROBABILITY TREES AND CONDITIONAL EXPECTATIONS
READING 5: PORTFOLIO MATHEMATICS
READING 6: SIMULATION METHODS
READING 7: ESTIMATION AND INFERENCE
READING 8: HYPOTHESIS TESTING
READING 9: PARAMETIC AND NON-PARAMETIC TESTS OF INDEPENDENCE
READING 10: SIMPLE LINEAR REGRESSION
READING 11: INTRODUCTION TO BIG DATA TECHNIQUES

2
 READING 1: RATE AND RETURNS

 Learning outcomes:

[Link] RATES AND TIME VALUE OF MONEY

-Compound interest or interest on interest: the growth in the value of the investment
from period to period reflects not only the interest earned on the original principal
amount but also on the interest earned on the previous period’s interest earnings;
- Time value of investment: frequently call for determining the FV of an investment
CF as a result of effects of compound interest;
- Future value: projecting the cash flows forward, on the basis of an appropriate
compound interest rate, to the end of investment’s life;
- Present value: bringing the CF from investment back to the beginning of the
investment’s life based on an appropriate compound rate;
- Time line: a diagram of the CF associated with a TVM problem;
 Note:
- Cash flow at the end of one period is the same as the beginning of the next period.

- Required rate of return: is equilibrium IR for a particular investment, return that

investors and savers require to get them to lend their fund.

- Discount rate=Required rate of return

- Opportunity cost: interest rate of current consumption;

3
Determinants of Interest Rates

- Real risk-free rate: theoretical rate on single-period loan that has no expectation of
inflation in it, investor’s increase in purchasing power.

Nominal risk-free rate = real risk-free rate + expected inflation rate

- Other type of risk: each added to increase the required rate of return
+ Default risk: risk a borrower will not make a promised payments in a timely manner;
+ Liquidity risk: risk of receiving less than fair value for an investment if it must be sold
for cash quickly;
+ Maturity risk: prices of long-term bonds are more volatile than those short-term
bonds.
Longer maturity bonds have more maturity risk than short-term bonds and require a
maturity risk premium.
Required interest rate on a security = nominal risk free rate (T-bills) + default risk
premium + liquidity risk premium + maturity risk premium.
Many countries have short-term government debt whose interest rate can be
considered to represent the nominal risk-free interest rate over that time horizon in
that country
[Link] OF RETURN
Holding Period Return can be measured over a single period or over multiple periods
A holding period return, R, is the return earned from holding an asset for a single
specified period of time

A holding period return can be computed for a period longer than one year

Arithmetic or Mean Return

The arithmetic mean return assumes that the amount invested at the beginning of each
period is the same.
Geometric Mean Return
The geometric mean return to account for the compounding of return,

4
 A geometric mean return provides a more accurate representation of the growth in
portfolio value over a given time period than the arithmetic mean return

In general, the arithmetic return is biased upward unless each of the underlying holding
period returns are equal. The bias in arithmetic mean returns is particularly severe if holding
period returns are a mix of both positive and negative returns
The Harmonic Mean
-Harmonic mean: is used for certain computations, such as the average cost of shares purchased
over time.

Both the trimmed and the winsorized means seek to minimize

the impact of outliers in a dataset
- Trimmed mean (trung bình lược bỏ): computed by excluding a stated small percentage
of the lowest and highest values and then computing an arithmetic mean of the remaining values
- Winsorized mean (trung bình gán): It is calculated by assigning a stated percentage of
the lowest values equal to one specified low value and a stated percentage of the highest values
equal to one specified high value. E.x95% winsorized mean sets the bottom 2.5% of values equal
to the value at or below which 2.5% of all the values lie

5
 4. MONEY-WEIGHTED AND TIME-WEIGHTED RETURN
The money-weighted return accounts for the money invested and provides the
investor with information on the actual return she earns on her investment.
The money-weighted return and its calculation are similar to the internal rate of return
Amounts invested are cash outflows from the investor’s
perspective and amounts returned or withdrawn by the investor, or the money that
remains at the end of an investment cycle, is a cash inflow for the investor.

6
 Money-Weighted Return for a Dividend-Paying Stock
At time t = 0, an investor buys one share at a price of USD200
At time t = 1, he purchases an additional
share at a price of USD225
At the end of Year 2, t = 2, he sells both shares at a price
of USD235
During both years, the stock pays a dividend of USD5 per share

Time-Weighted Returns
An investment measure that is not sensitive to the additions and withdrawals of funds
is the time-weighted rate of return.
The time-weighted rate of return measures the
compound rate of growth of USD1 initially invested in the portfolio over a stated
measurement period.

7
 Daily valuation is commonplace. Suppose that a portfolio is valued daily over the course
of a year. To compute the time-weighted return for the year, we first compute each
day’s holding period return. We compute 365 such daily returns, denoted R1, R2, …,
R365. We obtain the annual return for the year by linking the daily holding period
returns in the following way: (1 + R1) × (1 + R2) × … × (1 + R365) – 1.

5. ANNUALIZED RETURN
The period during which a return is earned or computed can vary and often we have
to annualize a return that was calculated for a period that is shorter (or longer) than one year
Non-annual Compounding
Recall that interest may be paid semiannually, quarterly, monthly, or even daily. To
handle interest payments made more than once a year, we can modify the present
value formula as follows.

Annualizing Returns
To annualize any return for a period shorter than one year, the return for the periodmust
be compounded by the number of periods in a year.
A monthly return is compounded 12 times, a weekly return is compounded 52 times, and
a quarterly returnis compounded 4 times. Daily returns are normally compounded 365 times

8
 Continuously Compounded Returns
An important concept is the continuously compounded return associated with a holding
period return, such as R1
The continuously compounded return associated with a
holding period return is the natural logarithm of one plus that holding period return,
or equivalently, the natural logarithm of the ending price over the beginning price
(the price relative)

9
 6. OTHER MAJOR RETURN MEASURES AND THEIR
APPLICATIONS
Gross and Net Return
A gross return is the return earned by an asset manager prior to deductions for
management expenses, custodial fees, taxes, or any other expenses that are not directly related to
the generation of returns but rather related to the management and administration of an
investment.
Thus, gross return is an appropriate measure for evaluating and comparing the investment
skill of asset managers because it does not include any fees related to the management and
administration of an investment.
Net return is a measure of what the investment vehicle (e.g., mutual fund) has earned
for the [Link] return accounts for (i.e., deducts) all managerial and administrative expenses
that reduce an investor’s return.
Pre-Tax and After-Tax Nominal Return
Many investors are concerned about the possible tax liability associated with their
returns because taxes reduce the net return that they receive.
Capital gains and income may be taxed differently, depending on the jurisdiction. The
after tax nominal returns is computed as the total return minus any allowance for taxes on
dividends, interest, and realized gains
Real Returns

Real returns are particularly useful in comparing returns across time periods because
inflation rates may vary over time.
Leveraged Return
Using borrowed capital, debt, the size of the leveraged position increases by the
additional, borrowed capital

10
11
 READING 2: TIME VALUE OF MONEY IN FINANCE

2. TIME VALUE OF MONEY IN FIXED INCOME AND

EQUITY

Fixed-Income Instruments and the Time Value of Money

The discount rate for fixed-income instruments is an interest rate, and the rate of return
on a bond or loan is often referred to as its yield to maturity (YTM).

12
 The investor’s sole source of return is the difference between the price paid (PV) and
full principal (FV) received at maturity. This type of bond is often referred to as a
zero-coupon bond given the lack of intermediate interest cash flows, which for bonds
are generally referred to as coupons

13
Coupon Instrument
A periodic cash flow pattern for fixed-income interest payments with principal repaid
at maturity is shown in Exhibit 2. In this case, all the periodic cash flow payments are
identical and occur on a semiannual basis.

14
Annuity Instruments

Examples of fixed-income instruments with level payments, which combine interest

and principal cash flows through maturity, include fully amortizing loans such as
mortgages and a fixed-income stream of periodic cash inflows over a finite period
known as an annuity. Exhibit 3 illustrates an example of level monthly cash flows
based upon a mortgage

15
 Mortgage Cash Flows
An investor seeks a fixed-rate 30-year mortgage loan to finance 80 percent of
the purchase price of USD 1,000,000 for a residential building.
1. Calculate the investor’s monthly payment if the annual mortgage rate is 5.25
percent

16
Equity Instruments and the Time Value of Money
One way to value a company’s shares is by discounting
expected future cash flows using an expected rate of return (r). These cash flows
include any periodic dividends received plus the expected price received at the end
of an investment horizon.

17
18
19
Constant and Changing Dividend Growth

Recall that based on a constant GBP1.50 annual dividend and required return of 15
percent, we showed Shipline PLC’s expected stock price to be GBP10.00.
Suppose instead that an investment analyst assumes that Shipline will grow its
annual dividend by 6 percent per year indefinitely.
1. How does Shipline’s expected share price change under the analyst’s
constant growth assumption?

2. How does Shipline’s expected share price change if we assume instead an

initial dividend growth of 6 percent over a three-year period followed by
constant 2 percent dividend growth thereafter?

[Link] RETURN AND GROWTH

20
 In the case of a discount bond or instrument, recall that an investor receives a
single principal cash flow (FV) at maturity, with (FV - PV) representing the implied
return

The YTM assumes an investor expects to receive all promised cash flows through
maturity and reinvest any cash received at the same YTM

Equity Instruments, Implied Return, and Implied Growth

The price of a share of stock reflects not only the required return but also the growth of
cash flows

21
 Thus, the implied return on a stock given its expected dividend yield and growth is given
by Equation 21, as follows

Rather than comparing equity share prices directly in currency terms, a common practice
is to compare ratios of share price to earnings per share, or the
price-to-earnings ratio.

22
4. CASH FLOW ADDITIVITY

Under cash flow additivity, the present value of any future cash flow stream indexed at
the same point equals the sum of the present values of the cash flows.
Implied Forward Rates Using Cash Flow Additivity
Investment strategy 1: Invest today for two years at a known annualized yield
of 3.50 percent.
Investment strategy 2: Invest today for one year at a known yield of 2.50
percent and reinvest in one year’s time for one year at a rate of F1,1 (the one year
forward rate starting in one year).

23
Forward Exchange Rates Using No Arbitrage

24
Option Pricing Using Cash Flow Additivity
Let’s assume an asset has a current price of 40 Chinese yuan (i.e., CNY40). The asset
is risky in that its price may rise 40 percent to CNY56 during the next time period or
its price may fall 20 percent to CNY32 during the next time period.

■Price Increase: The 40 percent increase results in a contract value of CNY 6

(=CNY56 - CNY50), as the contract owner will choose to buy the asset at
CNY50 and is able to sell it at a market price of CNY56.

■Price Decrease: The 20 percent decrease results in a contract value of zero

(CNY0), as the contract owner will choose not to buy the asset at CNY50
when the market price is only CNY32

25
For example, assume at t= 0 an investor creates a portfolio in which the contract is sold at a price
of co and 0.25 units of the underlying asset are purchased. This portfolio is called a replicating
portfolio in that it is designed specifically to create a matching future cash flow stream to that of
a risk-free asset. Denoting the value of the portfolio as V at t = 0 and at t= 1 under both the price
increase and price decrease scenarios, we have the following:

26
 READING 3: Statistical Measures of Asset Returns

A measure of central tendency specifies where the data are centered. For
a return series, a measure of central tendency shows where the empirical distribution
of returns is centered, essentially a measure of the “expected” return based on the
observed sample. Measures of location, mean, the median, and the mode include
not only measures of central tendency but also other measures that illustrate other
aspects of the location or distribution of the data.
The Arithmetic Mean
The arithmetic mean is the sum of the values of the
observations in a dataset divided by the number of observations
- Sample mean:

Population mean and sample mean are both examples of arithmetic mean.
-Median: midpoint of data set when data is arranged in ascending or descending order.
Example: What is the median return for five portfolio managers with 10 year annualized
total records of: 30%, 15%, 25%, 21% and 23%.

Suppose we add a sixth manager to the previous example with a return of 28%. What is
the median return?
-Mode: value that occurs most frequently in data set. It may have more than one mode or even
no mode.
+ Unimodal: one value appears most frequently

27
 + Bimodal and trimodals: set of data has two or three values
Example: The mode
What is the mode of the following data set?
Data set: 30%, 28%, 25%, 23%, 28%, 15% and 5%.

Dealing with Outliers

In practice, although an extreme value or outlier in a financial dataset may represent
a rare value in the population, it may also reflect an error in recording the value of
an observation or an observation generated from a different population.

The first option is appropriate if the values are legitimate, correct observations,
and it is important to reflect the whole of the sample distribution
The second option excludes the extreme observations. One measure of central
tendency in this case is the trimmed mean, which is computing an arithmetic mean
after excluding a stated small percentage of the lowest and highest values. For example, a 5
percent trimmed mean discards the lowest 2.5 percent and the highest 2.5
percent of values and computes the mean of the remaining 95 percent of values.
The third option involves substituting values for the extreme values. A measure of
central tendency in this case is the winsorized mean. It is calculated assigning one specified low
value to a stated percentage of the lowest values in the dataset and
one specified high value to a stated percentage of the highest values in the dataset
Often comparing the statistical measures of datasets with outliers included and with
outliers excluded can reveal important insights about the dataset. Such comparison
can be particularly helpful when investors analyze the behavior of asset returns and
rate, price, spread and volume changes

28
 Measures of Location
-Quantile (known as measure of location): value at or below which stated proportion of the data
in a distribution lies.
+ Quartiles-the distribution is divided into quarters;
+ Quintiles-the distribution is divided into fifths;
+ Decile-the distribution is divided into tenths;
+ Percentile-the distribution is divided into hundredths (percent);

-When the location, Ly, is a whole number, the location corresponds to an actual observation.
-When Ly is not a whole number or integer, Ly lies between the two closest integer numbers
(one above and one below), and we use linear interpolation between those two places to
determine Py
The interquartile range (IQR) is the difference between the third quartile and the first quartile

Investment analysts use quantiles every day to rank performance—for example, the performance
of portfolios. The performance of investment managers is often characterized in terms of the
quartile in which they fall relative to the performance of their peer group of managers.
A box and whisker plot consists of a “box” with “whiskers” connected to the box. the
way to visualize the dispersion of data across quartiles is to use a diagram, such as a box and
whisker chart.
The “box” represents the lower bound of the second quartile and the upper bound of the
third quartile, with the median or arithmetic average noted as a measure of central tendency of
the entire distribution
The whiskers are the lines that run from the box and are bounded by the “fences,” which
represent the lowest and highest values of the distribution.

29
The first quartile corresponds to the lowest 25% of the ranked data, or −0.293% (and lower).

The third quartile corresponds to the lowest 75% of the ranked data, or 0.460% (and lower).

The interquartile range is the difference between the third and first quartiles, 0.460% and
−0.293%, or 0.753%.

Another form of box and whisker plot typically uses 1.5 times the interquartile range for the
fences. Thus, the upper fence is 1.5 times the interquartile range added to the upper bound of Q3,
and the lower fence is 1.5 times the interquartile range subtracted from the lower bound of Q2.

30
 [Link] OF DISPERSION
Dispersion is the variability around the
central tendency. If mean return addresses reward, then dispersion addresses risk
and uncertainty.
Dispersion is defined as the variability around the central tendency.
+ Central tendency: measure of the reward;
+ Dispersion: measure of risk;
-Range: is the distance between the largest and the smallest value (maximum value –
minimum value);

-Mean absolute deviation (MAD):average of the absolute values of the deviations of

individual observations from the arithmetic mean.

31
 -Population variance: average of the squared deviations from the mean;

-Population standard deviation: is the square root of the population variance and is
calculated as follow:

-Sample variance: s2 measure of dispersion applying when we are evaluating a sample of

n observation from a population.

The quantity n − 1 is also known as the number of degrees of freedom in estimating the
population variance
-Sample standard deviation: calculated by taking the square root of the sample variance.

32
Downside risk consider only outcomes less than mean (or some specific value) in calculating
risk measure.

Coefficient of variation

33
 4. Measure of shape of distribution:
-Skewness: distribution is not symmetrical, may be either positively or negatively skewed and
result from the occurrence of outliers in the data set.
(Outliers: observations with extraordinary large values, either positive or negative.)

+ Positively skewed: distribution with many outliers in the upper region, or right tail.
positive skew has frequent small losses and a few extreme gains

+ For unimodal distribution,

+ Negative skewed: distribution has a disproportionately large amount of

outliers that fall within its lower tail. skewed distribution shown has a long tail on its left
side

-A symmetric distribution has skewness of 0;

- A positively skewed distribution has positive skewness;
- A negatively skewed distribution has negative skewness;
LOS 2.m: Interpret kurtosis.
-Kurtosis: measure of combined weight of the tails of distribution relative to the rest of
distribution – that is, proportion of the total probability in the tail;
+ Leptokurtic:a distribution has fatter tails than normal distribution tends to generate
more-frequent extremely large deviation from the mean than normal distribution;
+ Platykurtic: a distribution has thinner tails than normal distribution;

34
 + Mesokurtic: same kurtosis as a normal distribution;

+Excess kurtosis: characterizes kurtosis relative to the normal distribution, is defined as

kurtosis minus three. Normal distribution, computed kurtosis is three;
(Normal distribution: excess kurtosis equal 0; leptokurtic: excess kurtosis greater than 0;
platykurtosis: excess kurtosis less than 0)
Most risk managers focus more on the distribution of returns in the tails of the
distribution. General, greater positive kurtosis and negative skew indicates increased risk.
Measures of sample skew and kurtosis:
-Sample skewness: equal to the sum of the cubed deviations from the mean divided by
the cubed standard deviations and number of observations.

When a distribution is right skewed, sample skewness is positive;

When a distribution is left skewed, sample skewness is negative;
Value of sample skewness in excess of 0.5 in absolute value are considered significant.
-Sample kurtosis: is measured using deviations raised the fourth power:

- Interpret kurtosis: measured relative to the kurtosis of a normal distribution, which is

3.
+ leptokurtic: positive value of excess kurtosis;
+ platykurtic: negative value of excess kurtosis;

35
 + excess kurtosis = sample kurtosis – 3;
Excess kurtosis >1 in absolute value are considered large;
Most equity return series have been found to be leptokurtic. If a return distribution has
positive excess kurtosis (leptokurtosis) and we use statistical models that do not account for the
fatter tails, we will underestimate the likelihood of very bad or very good outcomes

5. CORRELATION BETWEEN TWO VARIABLES

Scatter plots are method for displaying the relationship between two variables with one
variable on the vertical axis and the other on the horizontal axis, their paired observations can be
plotted as single point.

Correlation is a measure of the linear relationship between two random variables.

The sample covariance (sXY) is a measure of how two variables in a sample move together

covariance is a measure of the joint variability of two random variables.

- If the random variables vary in the same direction—for example, X tends to beabove its mean
when Y is above its mean, and X tends to be below its mean when Y is below its mean—then
their covariance is positive. If the variables vary in the opposite direction relative to their
respective means, then their covariance is negative.
- Definition of Sample Correlation Coefficient.

36
 The sample correlation coefficient is a standardized measure of how two variables in a
sample move together. The sample correlation coefficient (rXY) is the ratio of the sample
covariance to the product of the two variables’ standard deviations:

-
Limitations of Correlation Analysis.
Correlation measures the linear association between two variables, but it may not always
be reliable. Two variables can have a strong nonlinear relation and still have a very low
correlation.
For example, the relation Y= (X− 4)2is a nonlinear relation contrasted to the linear
relation Y= 2X− 4. Even though these two variables are perfectly associated, there is no linear
association between them
Role of outliers (extreme values) in the correlation of two variables; If removing the
outliers significantly reduces the calculated correlation, further inquiry is necessary into whether
the outliers provide information or are caused by noise (randomness) in the data used
- Spurious correlation refers to correlation that is either the result of chance relationships
in a particular data set or present due to changes in both variables over time that is caused by
their association with a third variable.

37
 READING 4: Probability Trees and Conditional Expectations

1. Expected value and variance

-Expected value: weighted average of the possible outcomes for the variable.

-Expected variance: The variance is calculated as the probability-weighted sum of the

squared differences between each possible outcome and expected EPS

38
[Link] trees and conditional expectations
In investments, we make use of any relevant information available in making our
forecasts. When we refine our expectations or forecasts, we are typically updating them based on
new information or events; in these cases, we are using conditional expected values. The
expected value of a random variable X given an event or scenario
S is denoted E(X | S). Suppose the random variable X can take on any one of n distinct outcomes
X1, X2, …,Xn (these outcomes form a set of mutually exclusive and exhaustive events).
The expected value of X conditional on S is the first outcome, X1, times the probability of
the first outcome given S, P(X1 | S), plus the second outcome, X2 times the probability of the
second outcome given S, P(X2 | S), and so forth, as follows:

Total Probability Rule for Expected Value

The formula follows:
E(X) = E(X | S)P(S) + E(X | SC)P(SC), (5)
where SC is the “complement of S,” which means event or scenario “S” does not occur.
E(X) = E(X | S1)P(S1) + E(X | S2)P(S2) + … + E(X | Sn)P(Sn), (6)
Where S1, S2, …, Sn are mutually exclusive and exhaustive scenarios or events.

[Link]' FORMULA AND UPDATING PROBABILITY ESTIMATES

39
 -Bayes’ formula: update a given set of prior probabilities for a given event in response to
the arrival of new information:

Example: Bayes’ formula

40
 READING 5: Portfolio Mathematics

[Link] EXPECTED RETURN AND VARIANCE OF RETURN

Portfolio Expected Return Given a portfolio with n securities, the expected return on the
portfolio (E(Rp)) is a weighted average of the expected returns (R1 to Rn) on the component
securities using their respective weights (w1 to wn):

-Covariance: measure how two assets move together;

+ The covariance of RA with itself is equal to the variance of RA;

+ The covariance measures how one random variable moves with another random
variable, it may range from negative infinity to positive infinity;
+ Covariance of return is negative, when the return on one asset is above expected value,
the return on the other asset below its expected value;
+ Covariance of return is positive, returns on the both assets tend to be on the same side;

41
 There are n variance terms and n(n -1)/2 unique covariance terms

-Correlation coefficient:

42
2. FORECASTING CORRELATION OF RETURNS: COVARIANCE GIVEN A
JOINT PROBABILITY FUNCTION

3.

4.

43
4. PORTFOLIO RISK MEASURES: APPLICATIONS OF THE NORMAL
DISTRIBUTION
-Short fall risk: the probability that a port value or return will fall below a particular (target)
value or return over a given time period;
-Roy’s safety-first criterion: the optimal port minimizes the prob that return of the port
falls below some minimum acceptable level.
- Threshold level: minimum acceptable level;
- Roy’s safety-first criterion can be stated as:

- Formula for Roy’s safety-first criterion:

44
Two mainstays in managing financial risk are value at risk (VaR) and stress testing/scenario
analysis.
Stress testing and scenario analysis refer to a set of techniques for estimating losses in
extremely unfavorable combinations of events or scenarios.
Value at risk (VaR) is a money measure of the minimum value of losses expected over a
specified time period (e.g., a day, a quarter, or a year) at a given level of probability (often 0.05
or 0.01). Suppose we specify a one-day time horizon and a
level of probability of 0.05, which would be called a 95 percent one-day VaR. If this
VaR equaled EUR5 million for a portfolio, there would be a 0.05 probability that the
portfolio would lose EUR5 million or more in a single day (assuming our assumptions
were correct).

45
 READING 6: SIMULATION METHODS

2. LOGNORMAL DISTRIBUTION AND CONTINUOUS COMPOUNDING

-Lognormal distribution: is generated by the function ex, where x is normally distributed.

+ The lognormal distribution is skewed to the right

+ The lognormal distribution is bounded from below by zero so that it is useful for
modeling asset prices which never takes negative values;
+ A price relative is just the end-of-period price of the asset divided by the beginning
price (S1/S0) and is equal to (1 + the holding period return);
+ To get the end-of-period asset price, we can simply multiply the price relative times the
beginning-of-period asset price
-Continuous compounding: the effective annual rate based on continuous compounding for a
stated annual rate of Rcc
+ Effective annual rate = eRcc -1

46
 - Compounded rates of return are additive for multiple periods;

In general, the holding period return after T years, when the annual continuously compounded
rate is Rcc, is given by

Given investment results over a 2-year period, we can calculate the 2-year continuously
compounded return and divide by two to get the annual rate. Consider an investment that
appreciated from $1,000 to $1,221.40 over a 2-year period. The 2-year continuously
compounded rate is ln(1,221.40 / 1,000) = 20%, and the annual continuously compounded rate
(Rcc) is 20% / 2 = 10%.

3. Monte Carlo Simulation

-Monte Carlo simulation: is a technique based on the repeated generation of one or more
risk factors that affect security values to generate a distribution of security values.

- The limitation of MC:

+ fairly complex;
+ answer depends on the distribution of the risk factors and pricing/valuation model
used;
+ it is statistical not analytic (cause-and-effect relationships); (kết quả dựa trên phương
pháp thống kê, không phải kết quả chính xác)
+ cannot provide the insights that analytic methods can; (không phải phương pháp phân
tích và không thể kết luận về mối quan hệ nhân quả giữa các biến)
[Link]:
-Resampling, which repeatedly draws samples from the original observed data sample
for the statistical inference of population parameters.

47
- Bootstrap, one of the most popular resampling methods, uses computer simulation or statistical
inference without using an analytical formula such as a z-statistic or t-statistic.
Repeatedly draw samples from the original sample, and each resample is of the same size
a the original sample. Note that each item drawn is replaced for the next draw (i.e., the
identical element is put back into the group so that it can be drawn more than once)

48
 READING 7: ESTIMATION AND INFERENCE

Sampling methods
Probability sampling (chọn mẫu ngẫu nhiên) refers to selecting a sample when we know the
probability of each sample member in the overall population.
+ With random sampling, each item is assumed to have the same probability of being selected
+ simple random sampling a population of data and select our sample by using a computer to
randomly select a number of observations from the population, each data point has an equal
probability of being selected
Non-probability sampling (chọn mẫu phi ngâu nhiễn) is based on either low cost and easy
access to some data items, or on using the judgment of the researcher in selecting specific data
items. Less randomness in selection may lead to greater sampling error.
-Sampling error: is the difference between a sample statistic (the mean, variance, or
standard deviation of the sample) and its corresponding population parameter (the true mean,
variance, or standard deviation of the population)

-Simple random sampling: method of selecting a sample in such a way that each item or
person in the population being studied has the same likelihood of being included in the sample.
-Stratified random sampling: use a classification system to separate the population into
smaller groups based on one or more distinguished characteristics.
+ From each subgroup, or stratum, a random sample is taken and the results are pooled.
+ The size of the samples from each stratum is based on the size of the stratum relative to
the population.

49
 (nhóm màu khác nhau: trắng, đỏ, hồng)
- Cluster sampling:
Division or classification of the population into subpopulation groups, called clusters. In
this method, the population is divided into clusters, each of which is essentially a mini-
representation of the entire populations. Then certain clusters are chosen as a whole using
simple random sampling.
If all the members in each sampled cluster are sampled, this sample plan is referred to as
one-stage cluster sampling. (lựa chọn ngẫu nhiên các cụm và chọn hết tất cả các cá thể trong
cụm đó)

If a subsample is randomly selected from each selected cluster, then the plan is referred
as two-stage cluster sampling. (chúng ta sẽ lựa chọn ngẫu nhiên các cụm, sau đó, trong mỗi
cụm, chúng ta sẽ lựa chọn ngẫu nhiên các cá thể trong mỗi cụm)

Phương pháp chọn mẫu này có ưu điểm là tổ chức gọn nhẹ, giảm được chi phí (time
efficient, cost efficient) nhưng nếu số đơn vị được chọn chỉ tập trung vào một số khối thì có thể
dẫn đến sai số lớn khi giữa các khối có sự khác biệt nhau nhiều. Cách chọn này thường áp dụng
trong điều tra chất lượng sản phẩm mà khi sản xuất xong, sản phẩm đã được đóng kiện. Các đơn
vị trong một khối có sự khác nhau đáng kể song giữa các khối lại giống nhau về bản chất.

A major difference between cluster and stratified random samples is that in cluster
sampling, a whole cluster is regarded as a sampling unit and only sampled clusters are
included.
In stratified random sampling, however, all the strata are included and only
specific elements within each stratum are then selected as sampling units.

50
 - Non-probability sampling methods (chọnmẫu phi ngẫunhiên)
+ Convenience sampling (lấy mẫu thuận tiện) refers to selecting sample data that based
on its ease of access, using data that are readily available. Because such a sample is typically not
random, sampling error will be greater. It is most appropriate for an initial look at the data prior
to adopting a sampling method with less sampling error. (companies distribute their promotional
pamphlets and ask questions at a mall or on a crowded street with randomly selected
participants)
Because the samples are selected conveniently, they are not necessarily representative of
the entire population, and hence the level of the sampling accuracy could be limited.
But the advantage of convenience sampling is that data can be collected quickly at a low
cost.
+Judgmental Sampling: This sampling process involves selectively handpicking
elements from the population based on a researcher’s knowledge and professional judgment
judgmental sampling (mẫu xét đoán) could be affected by the bias of the researcher
and might lead to skewed results that do not represent the whole population
the specialty of researchers is critical to select a more representative sample than by using
other probability or non-probability sampling methods, judgmental sampling allows researchers
to go directly to the target population of interest
For example, study which needed to know what it takes to graduate college, the only
people can give the reseacher first hand advice are the individuals who graduted college.
A Sampling Distribution: of sample statistic is a prob distribution of all possible sample
statistics computed from a set of equal-size samples that were randomly drawn from the sample
population.

51
CENTRAL LIMIT THEOREM AND INFERENCE

Important properties of the central limit theorem:

+ If the sample size n is sufficiently large (n≥30), the sampling distribution of the sample
mean will approximately normal.
+ The mean of the population, μ, and the mean of the distribution of all possible sample
means are equal.
+ The variance of the distribution of sample mean is σ2/n, the population variance divided
by the sample size.
-Standard error of the sample mean: standard distribution of the distribution of the
sample means.

When do not know the population standard deviation and need to use sample standard
deviation

52
BOOTSTRAPPING AND EMPIRICAL SAMPLING DISTRIBUTIONS
-Resampling, which repeatedly draws samples from the original observed data sample
for the statistical inference of population parameters.
-Jackknife is another resampling technique for statistical inference of population
parameters
Jackknife samples are selected by taking the original observed data sample and leaving
out one observation at a time from the set (and not replacing it).
Jackknife method is often used to reduce the bias of an estimator, and other applications
include finding the standard error and confidence interval of an estimator.

53
 For a sample of size n, jackknife usually requires n repetitions

- Bootstrap, one of the most popular resampling methods, uses computer simulation or
statistical inference without using an analytical formula such as a z-statistic or t-statistic.
Repeatedly draw samples from the original sample, and each resample is of the same size
a the original sample. Note that each item drawn is replaced for the next draw (i.e., the
identical element is put back into the group so that it can be drawn more than once)

54
55
 READING 8: Hypothesis Testing

HYPOTHESIS TESTS FOR FINANCE

-Hypothesis is a statement about the value of a population parameter developed for the
purpose of testing a theory or belief.
+ Hypotheses are stated in terms of the population parameter to be tested, like the
population mean.
E.x Hypothesis iss the mean daily
+ Hypothesis testing procedure:

The Null Hypothesis and Alternative Hypothesis:

- The null hypothesis: is the hypothesis that the researcher wants to reject.

56
 -The alternative hypothesis: what is concluded if there is sufficient evidence to reject the
null hypothesis. It is the hypothesis that are really trying to assess.
-can be one-sided or two-sided.
+ One-sided is referred to as a one-tailed test;
+ Two-sided test is referred to as a two-tailed test;
- Two-tailed test for the population mean may be structured as:

- Two-tailed test uses two critical values, the general decision is:
Reject Ho if: test statistic > upper critical value or test statistic < lower critical value.

-One-tailed hypothesis test of the population mean, the null and alternative hypotheses
are either:

The appropriate of hypotheses depends on whether we believe the population mean, μ, to

be greater than (upper tail) or less than (lower tail) the hypothesized value, μ0.
–z0.05= –1.645 corresponds to a cumulative probability equal to 5%, and the z 0.05=1.645
corresponds to a cumulative probability of 95%

57
 -Hypothesis testing involvestwo statistics: the test statistic calculated from sample data
and the critical value of the test statistic.
-A test statistic: is calculated by comparing the point estimate of the population
parameter with hypothesized value of parameter.

+ Population standard deviation is known:

+ Population standard deviation is not known:

Type I and Type II error:

Keep in mind that hypothesis testing is used to make inferences about the parameters ofa
given population on the basis of statistics computed for a sample that is drawn fromthat
population.
Must be aware that there is some probability that the sample, insome way, does not
represent the population, and any conclusion based on the sampleabout the population may be
made in error
When drawing inferences from a hypothesis test, there are two types of error:
+ Type I: the rejection of null hypothesis when it is actually true;
+ Type II: the failure to reject the null hypothesis when it is false;
The significance level is the probability of making a Type I error.

58
LOS 6.d: Explain a decision rule, the power of a test, and the relation between confidence
intervals and hypothesis tests.
-The decision rule: either to reject the null hypothesis or fail to reject the null hypothesis.
+ Determine a one-tailed or two-tailed hypothesis test;
+ Determine the distribution of the test statistic;
+ Determine critical value;
+ If the test statistic is greater than the value X, reject the null.
-The power of the test: the probability of correctly rejecting the null hypothesis when it is
false or 1 – P(type II error).

Decreasing the significance level from 5% to 1 %, will increase the probability of failing
to reject the false null (Type II error), reduce the power of the test.
- For given of sample size: increase the power of the test with the cost that the
probability of rejecting the true null increases (Type 1 error);
- For given significance level: decrease the probability of type II error and increase the
power of the test, by increasing sample size.
- The p-value is the probability of obtaining a test statistic leading to a rejection of the
null hypothesis, assuming the null hypothesis is true.
+ It is the smallest level of significance for which the null hypothesis can be rejected.

59
The Relation between confidence intervals and Hypothesis test:
- Confidence interval: a range of values within which the researcher believes the true
population parameter will lie
+ Formula to determine confidence interval:

Or

The range within which we fail to reject the null for a two-tailed hypothesis test at a
given significance level.

60
 -A pooled variance is used with the t-test for testing the hypothesis that the means of two
normally distributed populations are equal, when the variances of the populations are unknown
but assumed to be equal.
Assuming independent samples, the t-statistic in this case is computed as:

The t-test for equality of population means when the populations are normally distributed
and have variances that are unknown and assumed to be unequalusesthe sample variances
for both populations.
Assuming independent samples, the t-statistic in this case is computed as follows:

61
Test Concerning Differences between Means with Dependent Samples
“paired comparison”:test of whether the means of the differences between
observationsfor the two samples are different.
-If the observations in the two samples both depend on some other factor, using “paired
comparison”
+ “paired comparison”:test of whether the means of the differences between
observationsfor the two samples are different,
+ “paired comparison”:requires that the sample data be normallydistributed.

The general form of the test for any hypothesized mean difference, μdz, is as follows:

n=number of paired observations;

62
Test Concerning the Equality of Two Variances
-The chi-square test is used for hypothesis tests concerning the variance of a normally
distributed population.
- The hypotheses for a two-tailed test of a single population variance are structured as:

The chi-square distribution is asymmetrical and approaches the normal distribution in

shape as the degrees of freedom increase.

63
 The chi-square test compares the test statistic, ,to a critical chi-square value at a
given level of significance and n − 1 degrees of freedom.

Testing the equality of the variances of two normally distributed populations, based on two
independent random samples

64
Independent random samples:

-The hypotheses concerned with the equality of the variances of two populations are
tested with an F-distributed test statistic.

-The F-test is used under the assumption that the populations from which samples are
drawn are normally distributed and that the samples are independent.

- A convention is to use larger of two sample variances in the numerator to reduces the
number of F table needed.

-The hypotheses for the two-tailed F-test of differences in the variances canbe structured
as:

-The F-distribution is right-skewed and is bounded by zero on the left-hand side;

-The shape of the F-distribution is determined by two separate degrees of freedom, the
numerator degrees of freedom, df1and the denominator degrees of freedom, df2.
A convention is to use the larger of the two sample variances in the numerator to
reduce the number of F-tables needed.

65
 - The upper critical value is always greater than one;
- The lower critical value is always less than one;

PARAMETRIC VERSUS NONPARAMETRIC TESTS

Parametric tests (Thống kê tham số) rely on assumptions regarding the distribution of the
population andare specific to population parameters (Mean and variances).

66
 -Nonparametric tests (Thống kê không tham số) either do not consider a particular
population parameter or have few assumptions about the population that is sampled.
- Nonparametric tests are used:
+ The assumptions of parametric tests can’t be supported;
+ Second, whereas the underlying distribution of the population may be normal,
there may be extreme values or outliers that influence the parametric statistics but
not the nonparametric statistics. For example, we may want to use a nonparametric
test of the median, in the case of outliers, instead of a test of the mean;
+ Third, we may have a sample in which observations are ranked. In those cases, we
also use nonparametric tests because parametric tests generally require a stronger
measurement scale than ranks. For example, if our data were the rankings of investment
managers, we would use nonparametric procedures to test the hypotheses
concerning those rankings.
+ A fourth situation in which we use nonparametric procedures occurs when our
question does not concern a parameter. For example, if the question concerns whether
a sample is random or not, we use the appropriate nonparametric test (a “runs test”).
The nonparametric runs test is used to test whether stock price changes can be used
to forecast future stock price changes—in other words, a test of the random-walk
theory.

67
68
 READING 9: PARAMETRIC AND NON-PARAMETRIC TESTS OF
INDEPENDENCE

[Link] CONCERNING CORRELATION

Parametric Test of a Correlation

The parametric pairwise correlation coefficient is often referred to as Pearson cor-
relation, the bivariate correlation, or simply the correlation.
The appropriate test statistic for the hypothesis that the population correlation equals zero, when
the two variables are normally distributed, is:

Non-Parametric Test of Correlation: The Spearman Rank Correlation

Coefficient
The population under consideration meaningfully departs from
normality, we can use a test based on the Spearman rank correlation coefficient,
rS it is calculated on the ranks of the two

69
variables (e.g., X and Y) within their respective samples. The calculation of rS requires
the following steps:
1. Rank the observations on X from largest to smallest. Assign the number 1 to
the observation with the largest value, the number 2 to the observation with
second largest value, and so on.
2. Calculate the difference, di, between the ranks for each pair of observations
on X and Y, and then calculate divvdc (the squared difference in ranks)

70
 The test of hypothesis for the Spearman rank correlation depends on whether the
sample is small or large (n > 30). For small samples, the researcher requires a specialized table of
critical values, but for large samples, we can conduct a t-test using
the test statistic, which is t-distributed with n - 2 degrees of freedom.
In this example, for a two-tailed test with a 5 percent significance level, the critical values for n -
2 = 35 - 2 = 33 degrees of freedom are ±2.0345.

TESTS OF INDEPENDENCE USING CONTINGENCY

TABLE DATA
Categorical or discrete data, we cannot use the methods that we have discussed up to this
point to test whether the classifications of such data are independent.
The classification of the investment type is discrete, so we cannot use correlation to
assess the relationship between size and investment type

This table is referred to as a contingency table or a two-way table (because there

are two classifications, or classes—size and investment type)

71
 Test whether there is a relationship between the size and investment
type, perform a test of independence using a nonparametric test statistic that
is chi-square distributed:

This test statistic has (r - 1)(c - 1) degrees of freedom, where r is the number of rows
and c is the number of columns.

If, on the one hand, the observed values are equal to the expected values, the calculated
test statistic would be zero. If, on the other hand, there are differences between the observed and
expected values,
These differences are squared, so the calculated chi-square statistic will be positive
The null hypothesis of independence between the two classes; the alternative hypothesis
of dependence. chi-square statistic as 32.08025 with (3 - 1) × (3 - 1) = 4 degrees of freedom and
a two-sided test with a 5% level of significance, the critical value is 11.143 reject the null
hypothesis if the calculated χ2 statistic is greater than 11.143

72
 Reading10: Simple Linear Regression

Simple linear regression is to explain the variation in a dependent variable in terms of

the variation in a single independent variable.
The dependent variable is the variable whose variation is explained by the independent
variable.
The independent variable is the variable used to explain the variation of the dependent
variable
Linear regression requires a number of assumptions

73
74
The estimated regression equation is Long-term inflation = –0.0008 + 0.3339 (Long-term money
supply growth)

In a linear regression, the fitted regression line passes through the point corresponding to the
means of the dependent and the independent variables.

INTERPRETING A REGRESSION COEFFICIENT

The intercept is an estimate of the dependent variable when the independent variable takes on a
value of zero
The estimated slope coefficient is interpreted as the change in the dependent variable for a 1-unit
change in the independent variable.

75
Explain the assumption underlying the simple linear regression model and describe
residuals and residual plots indicate if these assumptions may be violated.

Assumption 1: Linearity
If the relationship between the independent and dependent variables is nonlinear in the
parameters, estimating that relation with a simple linear regression model will produce invalid
results.
The model will be biased, because it will under- and overestimate the dependent variable
at certain points.

Assumption 2: Homoskedasticity (phương sai không đổi)

Heteroskedasticity (phương sai thayđổi)

76
 Assumption 3: Independence
Observations (Y and X pairs) are uncorrelated with one another, meaning they are
independent
If there is correlation between observations (that is, autocorrelation), they are not
independent and the residuals will be correlated.
The assumption that the residuals are uncorrelated across observations is also necessary
for correctly estimating the variances of the estimated parameters of b0 and b1
It is important to examine whether the residuals exhibit a pattern, suggesting a violation
of this assumption

There is a pattern. These residuals are correlated, specifically jumping up in Quarter 4 and then
falling back the subsequent quarter
Assumption 4: Normality
The assumption of normality requires that the residuals be normally distributed. This does not
mean that the dependent and independent variables must be normally distributed; it only means
that the residuals from the model are normally distributed.
With normally distributed residuals, we can test a particular hypothesis about a linear regression
model. For large sample sizes, we may be able to drop the assumption of normality by appealing
to the central limit theorem;
GOODNESS OF FIT AND HYPOTHESI TESTS

77
-Sum of square regression: phần dư bình phương
-Sum of quares error: bình phương sai số

78
R2 coefficient of estimation
 F Statistic
An F-test assesses how well a set of independent variables, as a group, explains the
variation in the dependent variable.

79
 In multiple regression, the F-statistic is used to test whether at leastone independent
variable in a set of independent variables explains a significant portion of the variation of the
dependent variable

The F-statistic in regression analysis is one sided, with the rejection region on the right
side, because we are interested in whether the variation in Y explained (the numerator)
is larger than the variation in Y unexplained (the denominator).

For simple linear regression, there is only one independent variable, so the F-test is
equivalent to a t-test for statistical significance of the slope coefficient:
Ho:b1=0 versus Ha:b1≠0
To determine whether b1 is statistically significant using the F-test, the calculated F
statistic is compared with the critical F-value, Fc, at the appropriate level of significance.
The degrees of freedom for the numerator and denominator with one independent
variable are:

80
This test statistic is t-distributed with n− k− 1

81
PREDICTING DEPENDENT VARIABLES AND FUNCTIONAL FORMS.
Predicted values are values of the dependent variable based on the estimasted regression
coefficients and prediction about the value of the independent variable. They are values that are
predicted by regression equation, given an estimate of the independent variable.

82
DESCRIBE DIFFERENT FUNCTIONAL FORMS OF SIMPLE LIEAR REGRESSIONS.
Not every set of independent and dependent variables has a linear relation. In fact,
we often see non-linear relationships in economic and financial data. E.x Revenues grow at a rate
of 15% per year for several years, but then the growth rate eventually declines to just
5% per year.
1 the log-lin model, in which the dependent variable is logarithmic but the independent variable
is linear;
2 the lin-log model, in which the dependent variable is linear but the independent variable is
logarithmic; and
3 the log-log model, where both the dependent and independent variables are in logarithmic
form
The Log-Lin Model
In the log-lin model, the dependent variable is in logarithmic form and the independent variable
is not, as follows: (the log-lin model is better fit)

The lin-Log model:

83
The lin-log model is similar to the log-lin model, but only the independent variable
is in logarithmic form: (the lin-log model is better fit than lin-lin model)

The Log-Log Model

The log-log model, in which both the dependent variable and the independent variable are linear
in their logarithmic forms, is also referred to as the double-log model.

Selecting the Correct Functional Form

The key to fitting the appropriate functional form of a simple linear regression is examining the
goodness of fit measures—the coefficient of determination (R2), the F-statistic, and the standard
error of the estimate (se)—as well as examining whether there are patterns in the residual.

84
85
 READING 11: INTRODUCTION TO BIG DATA TECHNIQUES

HOW IS FINTECH USED IN QUANTITATIVE

INVESTMENT ANALYSIS?
In its broadest sense, the term fintech generally refers to technology-driven innovation
occurring in the financial services industry. For our purposes, fintech refers to technological
innovation in the design and delivery of financial services and products.
Fintech has advanced into decision-making applications based on complex machine-learning
logic, in which computer programs are able to learn how to complete tasks over time.

Big Data
Big data has been in use since the late 1990s and refers to the vast amount of information
being generated by industry, governments, individuals, and electronic devices.
Big Data includes data generated from traditional sources—such as stock exchanges,
companies, and governments—as well as non-traditional data types, also known as alternative
data, arising from the use of electronic devices, social media, sensor networks, and company
exhaust (information generated in the normal course of doing business).
As the internet and the presence of such networked devices have grown, the use of non-
traditional data sources, or alternative data sources—including social media (posts, tweets, and
blogs), email and text communications, web traffic, online news sites, and other electronic
information sources—has risen.

86
ADVANCED ANALYTICAL TOOLS: ARTIFICIAL
INTELLIGENCE AND MACHINE LEARNING

87
 Overfittingthe data might discover “false” relationships or “unsubstantiated” patterns that
will lead to prediction errors and incorrect output forecasts. Overfitting occurs when the ML
model learns the input and target dataset too precisely. In such cases, the model has been
“overtrained” on the data and treats noise in the data as true parameters.
When a model has been underfitted, the ML model treats true parameters as if they are
noise and is not able to recognize relationships within the training data. In such cases, the model
could be too simplistic. Underfitted models typically will fail to fully discover patterns that
underlie the dataML can be divided broadly into three distinct classes of techniques: supervised
learning, unsupervised learning, and deep learning.
In supervised learning, computers learn to model relationships based on labeled
training data. In supervised learning, inputs and outputs are labeled, or identified, for
the algorithm. After learning how best to model relationships for the labeled data, the
trained algorithms are used to model or predict outcomes for new datasets
In unsupervised learning, computers are not given labeled data but instead
are given only data from which the algorithm seeks to describe the data and their
structure. For example, grouping companies into peer groups based on their characteristics rather
than using standard sector or country groupings is an application of
unsupervised learning techniques.
In deep learning (or deep learning nets), computers use neural networks, often
with many hidden layers, to perform multistage, non-linear data processing to identify patterns.
Deep learning can use supervised or unsupervised ML approaches. By
taking a layered or multistage approach to data analysis, deep learning develops an
understanding of simple concepts that informs analysis of more complex concepts.
[Link] BIG DATA WITH DATA SCIENCE
Data science can be defined as an interdisciplinary field that harnesses advances in
computer science (including ML), statistics, and other disciplines for the purpose
of extracting information from Big Data (or data in general).
Data Processing Methods

88
Data Visualization
Visualization refers to how the data will be formatted, displayed, and summarized in
graphical form.
Traditional structured data can be visualized using tables, charts, and trends, whereas
non-traditional unstructured data require new techniques of data visualization.
Multidimensional data analysis consisting of more than three variables requires additional
data visualization techniques—for example, adding color, shapes, and sizes to the 3D charts
Another valuable Big Data visualization technique that is applicable to textual data
is a “tag cloud,” in which words are sized and displayed on the basis of the frequency
of the word in the data file.
Text Analytics and Natural Language Processing
Text analytics involves the use of computer programs to analyze and derive meaning
typically from large, unstructured text- or voice-based datasets, such as company filings,
written reports, quarterly earnings calls, social media, email, internet postings, and
surveys
Natural language processing (NLP) is a field of research at the intersection of
computer science, AI, and linguistics that focuses on developing computer programs
to analyze and interpret human language.

89
90
91
92
93
94
95