B2_Additonal question bank

FRM PART – I – QUANTITATIVE ANALYSIS
Topic - 12
Fundamentals of Probability
Learning Objectives
Fundamentals of Probability
LO 12 a: Describe an event and an event space.

LO 12 b: Describe independent events and mutually exclusive events.
LO 12 c: Explain the difference between independent events and conditionally independent events.
LO 12 d: Calculate the probability of an event for a discrete probability function.
LO 12 e: Define and calculate a conditional probability.
LO 12 f: Distinguish between conditional and unconditional probabilities.
LO 12 g: Explain and apply Bayes’ rule.
Compiled by ulurn Page 76

QUESTIONS
1. Which of the following statements regarding properties of probabilities are correct?

1. Probabilities always lie between 0 and 1 inclusive.
2. Probabilities always add up to 1.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Which of the following statements regarding probability concepts is least likely correct?
A. The outcomes of a discrete random variable can be counted.
B. The possible values of a individual continuous random variable has a zero probability
associated with it.
C. Mutually exclusive events are if the events have no outcomes in common
D. The possible values under consideration for a continuous random variable are finite.
3. Consider the random variable X = outcome when a fair die is rolled. The probability
for the following probability function: P(2 ≤ X<6) is equal to:
1
A.
6
4
B.
6
3
C.
6
5
D.
6
4. Which of the following statements regarding the formula for A and B independent events are
correct?
1. P(AB) = P(A) × P(B)
2. P(AB) = P(B) + P(A)
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
5. Consider a deck of playing cards, the probability of randomly selecting either an ace or a heart
= P(Ace or Heart) is equal to:
16
A.
52
17
B.
52

1
C.
52
13
D.
52
6. Consider the following probability matrix:
What is the joint probability of an average market and declining stock prices?
A. 25%
B. 10%
C. 15%
D. 40%
7. Which of the following statements regarding the conditional and unconditional probabilities
are correct?
1. An unconditional probability is a form of probability that is concerned with a specific event
only.
2. An unconditional probability is form of probability takes into account the fact that some
other event has occurred which is likely to affect the chance of our event occurring.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
8. Consider a deck of playing cards, the probability of randomly selecting a card that is both an
ace and a heart = P(Ace and Heart) is equal to:
16
A.
52
17
B.
52
1
C.
52
13
D.
52
9. Which of the following statements regarding Bayes’ formula are correct?

I. Bayes’ formula is used when we know that the event whose probability we have
just calculated has occurred, and we wish to evaluate conditional probabilities based on
this fact.

II. Bayes’ formula is used to update our knowledge of a specific event occurring in the
light of new information received, i.e. that an event has occurred.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
10. Which of the following formulas cannot be used to calculate updated probabilities?
11. An analyst has developed a ratio to identify company’s expectation to experience declining PE
multiples over time. Research shows that 55% of firms with declining PE’s have a negative
ratio, while only 25% of firms not experiencing a decline in PE’s have a negative ratio. The
analyst expects that 15% of all publicity traded companies will experience a decline in PE next
year. The analyst randomly selects a company, and its ratio is negative. Based on Bayes’
theorem, what is the probability that the company will experience a PE decline next year.
A. 28.00%
B. 38.97%
C. 8.25%
D. 99.40%
12. Matthew Nganou, FRM, is in the process of analyzing a selection of bond managers. The
following historical data is available:
The star bond managers are expected to outperform the market 80% of the time. The average
bond managers are expected to outperform the market 40% of the time. The probability of a
bond portfolio outperforming the market in any given year is independent of the previous
year’s performance. 30% of managers are considered to be stars and 70% are considered to be
average.
A fund manager that has worked in bonds for 5 years has outperformed the market each and
every year. What is the probability that the manager was a star when they started working as a
bond manager 5 years ago?
A. 30%
B. 70%
C. 0.24%
D. 32.77%

average.
A fund manager that has worked in bonds for 5 years has outperformed to market each and
every year. What is the probability that a star manager will outperform the market for 5 years
in a row?
A. 30%
B. 70%
C. 0.24%
D. 32.77%
average.
A fund manager that has worked in bonds for 5 years has outperformed to market each and
every year. What is the probability that a star manager will be a star manager today (i.e. after 5
years)
A. 30%
B. 6.80%
C. 32.77%
D. 93.27%
15. An insurance company estimates that 40% of policyholders who have only an auto policy will
renew next year, and 70% of policyholders who have only a homeowner policy will renew next
year. The company estimates that 80% of policyholders who have both an auto and a
homeowner policy will renew at least one of those policies next year. Company records show
that 70% of policyholders have an auto and a homeowner policy. Using the company’s
estimates, what is the percentage of policyholders that will renew at least one policy next year?
A. 29%
B. 41%
C. 53%
D. 57%

SOLUTIONS
1. C
Properties of probabilities
1. Probabilities always lie between 0 and 1 inclusive. This is a very important point to
remember.
2. Probabilities always add up to 1.
2. D
Discrete random variable – This is a random variable whose possible values can be listed, and
each value has some positive probability associated with it. The outcomes can be counted.
Continuous random variable – This is a random variable whose possible values cannot
be listed, as they are too numerous. There are normally many decimal places involved.
The possible values under consideration are infinite, and, as you will see later, each
individual value has zero probability associated with it. The outcomes can be measured.
Mutually exclusive – Two (or sometimes more) events are defined as being mutually exclusive
if the events have no outcomes in common.
3. B
X can take on values 1, 2, 3, 4, 5 and 6, and because the die is fair, each outcome has an equal
1
probability of occurring. This is an example of a discrete uniform random variable.
6
Calculating probabilities for such a random variable is simple.
Note that 2 is included because of the equal sign, but 6 is not.
4. A
This formula for A and B independent events is as follows:
P(AB) = P(A) × P(B) or alternatively P(AB) = P(B) × P(A)
The addition rule (statement 2) is concerned with the probability that at least one out of two
events occurs. This means that one OR both of the events can occur.
5. A
P(Ace or Heart) = P(A or B).
P(A or B) = P(A) + P(B) – P(AB)
This means that the chance of drawing a card at random from a pack of cards and
16 4
getting a card, which is either an Ace or a Heart or both is or .
52 13

6. B
7. C
Unconditional probability – This form of probability is concerned with a specific event only,
and does not take into account other events, which might occur simultaneously.
Conditional probability – This form of probability takes into account the fact that some other
event has occurred which is likely to affect the chance of our event occurring.
8. C
9. C
Bayes’ formula is used when we know that the event whose probability we have just calculated
has occurred, and we wish to evaluate conditional probabilities based on this fact.
Effectively, we are using Bayes’ formula to update our knowledge of a specific event occurring
in the light of new information received, i.e. that an event has occurred.
10. D
BAYES FORMULA:

11. A
X1 – PE will decline can be defined as: P(BX1) – ratio is negative when PE declines.
X2 – PE will go up can be defined as: P(BX2) – ratio is negative when PE goes up.
We are looking for the probability that PE will decline given a negative ratio.
Bayes’ theorem can be applied as follows:
12. A
Let us define the probabilities:
P of being a star = P(s) = 0.30
P of being average = P(a) = 0.70
Based on this, the unconditional probability that a random manager was a star five years ago is
30%.
13. D
Let us define the probabilities:
P of being a star = P(s) = 0.30
P of being average = P(a) = 0.70
P of outperforming the market = P(o)
In any single year, given that the probabilities are unconditional, the probability that a star will
outperform the market is 70%.
In other words:
P(o | s) = 0.80
To do this for five years in a row =
0.805 = 32.77% or 0.3277.
14. D
In order to work this out we need to use Bayes.
In other words, what is the probability of him being a star given that he has already
outperformed the market (for 5 years).
Step 1:
Let start of with the probability of him outperforming given that he is a star. This is equal to:
P(o | s) = 0.80

To do this for five years in a row =

0.805 = 32.77% or 0.3277
Step 2:
What is the probability that he is a star?
P(s) = 0.30
Step 3:
What is the probability that the manager will outperform =
P(o) = P(o | s) x P(s) + P(o | a) x P(a)
P(o) = 0.805 x 0.3 + 0.405 x 0.7
P(o) = 9.83% + 0.71%
P(o) = 10.54%
Now we can apply Bayes:
15. D
Let: A = event that a policyholder has an auto policy H = event that a policyholder has a
homeowner’s policy Then, based on the information given:

Topic - 13
Random Variables [QA-2]
Learning Objectives
Random Variables
LO 13 a: Describe and distinguish a probability mass function from a cumulative distribution

function, and explain the relationship between these two.
LO 13 b: Understand and apply the concept of a mathematical expectation of a random variable.
LO 13 c: Describe the four common population moments.
LO 13 d: Explain the differences between a probability mass function and a probability density
function.
LO 13 e: Characterize the quantile function and quantile-based estimators.
LO 13 f: Explain the effect of a linear transformation of a random variable on the mean, variance,
standard deviation, skewness, kurtosis, median, and interquartile range.

QUESTIONS
1. Which of the following statements regarding properties of the sample mean are correct?
I. The sample mean is unique.
II. The sample mean the mean calculated by adding up the values in your sample and
dividing by the sample size minus 1.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Consider the following data: 8, 7, 9, 10. What is the median of the data set?
A. 8
B. 7
C. 9.5
D. 8.5
3. Consider the following probability distribution of Revenue for Tradealot:

Revenue probability distribution:
The expected Revenue for Tradealot is:

A. 200
B. 190
C. 300
D. 100
4. Which of the following is not a property of expectations?

A. E[X + Y] ≠ E[X] + E[Y]
B. E[cX ] = cE[X]
C. E[XY] ≠ E[X] x E[Y]
D. E[XY] = E[X] x E[Y]
5. Which of the following statements regarding the PMF is not correct?

A. The PMF, is the probability of a random variable taking on a certain value.
B. The value returned by the PMF must be non-negative.
C. The sum of the values of the random variable must equal one.
D. The PMF is the cumulative or total probability.

6. Consider the returns of the following four different portfolios over the past year: 10%, 30%, 5%
and 15%.
What is the standard deviation of the portfolios returns?
A. 15.0%
B. 87.5%
C. 25.0
D. 9.3%
7. The second central moment – the variance – can be calculated using which of the following
formulas?
8. Which of the following statements are correct when dealing with kurtosis?
I. Kurtosis = the level of symmetry of the data.
II. Zero kurtosis = 3
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
9. Consider the following data set: 0, 1, 1, 1, 997.

Which of the following descriptions best describe this data set?
A. Negatively skewed
B. Positively skewed
C. Mesokurtic
D. Leptokurtic
10. The recent performance of Prudent Fund, with USD 50 million is assets, has been weak and the
institutional sales group is recommending that it be merged with Aggressive Fund, a USD 200
million fund. The returns on the Prudent Fund are normally distributed with a mean of 3% and
a standard deviation of 7%. The returns on Aggressive Fund are normally distributed with a
mean of 7% and a standard deviation of 15%. Senior management has asked an analyst to
estimate the likelihood that returns on the combined portfolio will exceed 26%. Assuming the
returns on the two funds are independent, the analyst’s estimate for the probability that the
returns on the combined fund will exceed 26% is closest to:
A. 1.0%
B. 2.5%
C. 5.0%
D. 10.0%

SOLUTIONS
1. D
Sample mean – This is the mean calculated by adding up the values in your sample and
dividing by the sample size.
Because there are many different possible samples, there are many possible values for the
means of the various samples. Thus, the sample mean is not unique.
2. D
We just take the median to be the average of the middle 2 values. In this case, the middle two
8  9
values are 8 and 9, and their average is  8.5 .
2
3. B
E(X) = P(xi)xi = P(x1)x1 + P(xn)xn
E(Revenue) = (0.30)100 + (0.50)200 +(0.20)300
E(Revenue) = 30 + 100 + 60
E(Revenue) = 190
4. A
Property 1
E[X + Y] = E[X] + E[Y]
Property 2
E[cX ] = cE[X]
Property 3
E[XY] ≠ E[X] x E[Y]
Property 4
E[XY] = E[X] x E[Y]
Property 5
E[X2] ≠ E[X]2
5. D
The first function, the PMF, is the probability of a random variable taking on a certain value. As
a result of the PMF returning a probability, it must contain the following two properties:
 The value returned by the PMF must be non-negative.
 The sum of the values of the random variable must equal one.
The second function is called the CDF which is a cumulative or total probability.

6. D
7. A
Variance is calculated as follows:
8. B
Skewness = The level of symmetry of the data.
Zero kurtosis = 3
9. B
The mean of the data = 200. The median is the middle value, so in this case, the median is 1.
Because the mean is much larger than the median, courtesy of the large outlier, we can describe
the data set as being positively skewed.
Effectively, the large outlier distorts the mean, pulling it over to the right, making the
distribution positively skewed.
10. C

Topic - 14
Common Univariate Random Variables [QA-3]
Learning Objectives
Common Univariate Random Variables
LO 14 a: Distinguish the key properties and identify the common occurrences of the following
distributions: uniform distribution, Bernoulli distribution, binomial distribution, Poisson
distribution, normal distribution, lognormal distribution, Chi-squared distribution, Student’s t, and
F-distributions.
LO 14 b: Describe a mixture distribution and explain the creation and characteristics of mixture
distributions.

QUESTIONS
1. Which of the following statements regarding properties of a uniform distribution are correct?
I. The uniform distribution graph is a straight line and thus yields uniform probabilities.
 X 2  X1 
II. The uniform distribution probability calculation is: P  X1  X  X 2  
b  a
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Which of the following statements regarding a random variable being binomial Is not correct?
A. The number of trials is fixed.
B. Each trial may have only two possible outcomes: success and failure.
C. The trials are independent of one another.
D. The sampling is done without replacement.
3. Assume that the number of defaults within a loan portfolio follows a Poisson process. The
expected number of defaults in a one-year period is 10.
What is the probability that exactly 8 defaults will occur in a one-year period?
A. 0.26%
B. 11.26%
C. 1.12%
D. 100%
4. Which of the following statements regarding the normal distribution are not correct?
A. The normal distribution is a discrete distribution.
B. The normal distribution has a kurtosis equal to 3.
C. The normal distribution has zero skewness.
D. The tails on a normal distribution extend indefinitely.
5. The heights of people in a city are normally distributed with mean 170 cm and standard
deviation 10 cm. What is the probability that a randomly chosen person from this population
has a height that is greater than 150 cm?
A. 0.0228
B. 0.9772
C. -0.9772
D. 0.8413
6. Which of the following statements regarding normal and lognormal distributions is correct?
A. The lognormal distribution is positively skewed.
B. The lognormal distribution has a lower bound of 0 and is also bounded from above.
C. The lognormal distribution is a discrete distribution.

D. The normal distribution is asymmetrical about its mean.
7. When testing the hypothesis of the population variance, which test statistic will be used?
A. A chi-square statistic
B. A F-statistic
C. A Z-statistic
D. A t-statistic
8. Which of the following statements regarding a mixture distribution are correct?

I. The skewness measure of the distribution can be changed by combining distributions with
different means.
II. The kurtosis measure of the distribution can be changed by combining distributions with
different variances.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
9. A portfolio manager holds three bonds in one of the portfolio and each bond has a 1-year
default probability of 15%. The event of default for each of the bonds is independent. What is
the probability of exactly two bonds defaulting over the next year?
A. 1.9%
B. 5.7%
C. 10.8%
D. 32.5%

SOLUTIONS
1. C
The uniform distribution graph is a straight line and thus yields uniform probabilities, hence
the name. The area under the graph between point a and point b equals 1.
In order to calculate the probability that x lies between two points, x1 and x2 say, we use:
2. D
There are a number of important conditions that need to apply for a random variable
to be binomial:
 The number of trials is fixed.
 Each trial may have only two possible outcomes: success and failure.
 The success probability p stays constant throughout the experiment and does not change
from trial to trial.
 The trials are independent of one another.
 The sampling is done with replacement.
3. B
4. A
The normal distribution is another special continuous distribution The graph is perfectly
symmetrical and therefore has zero skewness.
Its’ kurtosis is equal to 3. You will recall that excess kurtosis is measured relative to the number
3. We can, therefore, say that the graph has zero excess kurtosis.
The tails get smaller and smaller the further we move away from the mean. However, they
never equal zero – they extend indefinitely.
5. B
We want P(X>150). Standardizing gives P(Z>(150-170)/10) = P(Z>-2).
By symmetry, P(Z>-2) = P(Z<2). From tables, we can look up 2.00 and get 0.9772. So, the
answer is 0.9772.
6. A
One of the key differences between the two distributions: the normal distribution is
symmetrical about its mean, whereas the lognormal distribution is positively skewed.

The key properties of a lognormal distribution:

 It is positively skewed
 It has a lower bound of 0 but is not bounded from above
 It is a continuous distribution
7. A
When testing the hypothesis of the population variance, the test statistic to be used is a chi-
square (2) statistic with n-1 degrees of freedom.
8. C
The skewness measure of the distribution can be changed by combining distributions with
different means. The kurtosis measure of the distribution can be changed by combining
distributions with different variances.
9. B

Topic - 15
Multivariate Random Variables [QA - 4]
Learning Objectives
Multivariate Random Variables
LO 15 a: Explain how a probability matrix can be used to express a probability mass function.
LO 15 b: Compute the marginal and conditional distributions of a discrete bivariate random
variable.
LO 15 c: Explain how the expectation of a function is computed for a bivariate discrete random
variable.
LO 15 d: Define covariance and explain what it measures.
LO 15 e: Explain the relationship between the covariance and correlation of two random variables,
and how these are related to the independence of the two variables.
LO 15 f: Explain the effects of applying linear transformations on the covariance and correlation
between two random variables.
LO 15 g: Compute the variance of a weighted sum of two random variables.
LO 15 h: Compute the conditional expectation of a component of a bivariate random variable.
LO 15 i: Describe the features of an iid sequence of random variables.
LO 15 j: Explain how the iid property is helpful in computing the mean and variance of a sum of iid
random variables.

QUESTIONS
1. Consider the following data:

Variance of x = 100
Variance of y = 25
Covariance of x and y = 35
What is the value of the correlation coefficient?
A. 0.014
B. 0.70
C. 1.00
D. 0.28
2. Consider tow random variables X1 and X2 that are independent. What is the correlation
between X1 and X2?
A. 0
B. +1
C. -1
D. +0.5
3. Assume that X1 and X2 both have got univariate normal distributions, what can we conclude
about the joint distribution?
A. The joint distribution of X1 and X2 is bivariate normal.
B. The joint distribution of X1 and X2 is bivariate non-normal.
C. The joint distribution of X1 and X2 is not necessarily bivariate normal.
D. The joint distribution of X1 and X2 is univariate normal.
4. Consider the following probability matrix. The matrix describes the annual profits of two firms,
a big firm (X1) and a small firm (X2).
What is the marginal distribution of the big firm at the -$40m level?
A. 2.0%
B. 0%
C. 7.5%
D. 7.0%

Are the profits of Big firm and Small firm independent?

A. Yes
B. No
C. Unable to calculate
D. The profits are bivariate
What is the conditional distribution of Small firm’s if Big firm earns $100m for the year?
A. 0%, 22.58%, 41.94%, 35.48%
B. 0%, 3.5%, 6.5%, 5.5%
C. 0%, 22.58%, 41.94%, 38.48%
D. 0%, 22.58%, 45.94%, 35.48%

SOLUTIONS
1. B
R = 0.70
2. A
Independent random variables have a correlation of zero.
3. C
The joint distribution of X1 and X2 is not necessarily bivariate normal.
4. C
The marginal distributions are calculated by summing across the values at the $-40m level. In
other words, 2% + 4% + 1% + 0.5% = 7.5%.
5. B
No. The returns will be independent if the joint is the product of the marginals.
Let us look at the upper left cell, which is equal to 2.0%.
The marginal distribution for the upper left cell for Big firm = 2%+4%+1%+0.5% = 7.5%.
The marginal distribution for the upper left cell for Small firm = 2%+4%+1%+0.0%= 7.0%.
The product of the two = 7.5% x 7.0% = 0.52%
This is not equal to 2.0%, and as such the profits are not independent.
6. A
The conditional distribution is the row that corresponds to the $100m and is normalized to 1.
The row that corresponds to the $100m is 0%, 3.5%, 6.5%, 5.5% = 15.5%
Standardizing to 1 = 0%, 22.58%, 41.94%, 35.48% = 100%

Topic - 16
Sample Moments [QA - 5]
Learning Objectives
Sample Moments
LO 16 a: Estimate the mean, variance, and standard deviation using sample data.
LO 16 b: Explain the difference between a population moment and a sample moment.
LO 16 c: Distinguish between an estimator and an estimate.
LO 16 d: Describe the bias of an estimator and explain what the bias measures.
LO 16 e: Explain what is meant by the statement that the mean estimator is BLUE.
LO 16 f: Describe the consistency of an estimator and explain the usefulness of this concept.
LO 16 g: Explain how the Law of Large Numbers (LLN) and Central Limit Theorem (CLT) apply to
the sample mean.
LO 16 h: Estimate and interpret the skewness and kurtosis of a random variable.
LO 16 i: Use sample data to estimate quantiles, including the median.
LO 16 j: Estimate the mean of two variables and apply the CLT.
LO 16 k: Estimate the covariance and correlation between two random variables.
LO 16 l: Explain how coskewness and cokurtosis are related to skewness and kurtosis.

QUESTIONS
1. According to the central limit theorem: A random variable X with a mean  and variance 2
the sampling distribution of the sample mean x all possible samples of size n will be
approximately normally distributed, provided that n is sufficiently large. Which of the
following statements is correct?
I. The mean = 
2
II. The variance 
n
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Consider the following data:
What is the median of the above data observations?

A. 0.5050
B. 0.4600
C. 0.5500
D. 1.0000

SOLUTIONS
1. C
According to the central limit theorem: a random variable X with a mean  and variance 2 the
sampling distribution of the sample mean x of all possible samples of size n will be
approximately normally distributed, provided that n is sufficiently large.
The mean = 
2
The variance  .
n
2. A
The median is the midpoint, which in this case sits between the third and fourth observations.
0.46  0.55
The median   0.5050 .
2

Topic - 17
Hypothesis Testing [QA - 6]
Learning Objectives
Hypothesis Testing
LO 17 a: Construct an appropriate null hypothesis and alternative hypothesis and distinguish

between the two.
LO 17 b: Differentiate between a one-sided and a two-sided test and identify when to use each test.
LO 17 c: Explain the difference between Type I and Type II errors and how these relate to the size
and power of a test.
LO 17 d: Understand how a hypothesis test and a confidence interval are related.
LO 17 e: Explain what the p-value of a hypothesis test measures.
LO 17 f: Interpret the results of hypothesis tests with a specific level of confidence.
LO 17 g: Identify the steps to test a hypothesis about the difference between two population means.
LO 17 h: Explain the problem of multiple testing and how it can bias results.

QUESTIONS
1. Which of the following statements regarding the hypothesis test are not correct?
A. Ho: b1 ≥0
B. Ha: b1 < 0
C. Ho states that there is no relationship between the variables
D. A failure to reject Ho would indicate a statistically significant relationship between the
variables.
2. Whilst sampling to infer population parameters and using a two-tailed hypothesis testing
Megan Crane observes a t-statistic of 2.74 based on a sample of 21 observations where the
theoretical mean is 0. The null and alternative hypotheses are:
A. Ho: population mean = 0; Ha: population mean > 0
B. Ho: sample mean = 0; Ha: sample mean not equal to zero
C. Ho: population mean = 0; Ha: population mean not equal to 0
D. Ho: population mean = 1; Ha: population mean > 0
3. Two independent samples are taken and their statistics computed as per below:
Population A: Sample size 13, mean 340 and standard deviation 41
Population B: Sample size 13, mean 299 and standard deviation 46
A researcher hopes to demonstrate that the mean of population B is less than the mean of
population A to a 5% level of significance. The critical t-value assuming that the two population
variances are equal is nearest to:
A. 1.7109
B. 2.0639
C. 1.7056

D. 2.0555
4. An analyst believes that the average PE ratio for an index is at the most 12. She finds that the
average PE and standard deviation of a sample of 30 is 13 and 5 respectively. What is the value
of the test statistic?
A. 1.10
B. 6.00
C. 3.37
D. 12.0
5. Which of the below does not represent a step of hypothesis testing?

A. Making the economic or investment decision
B. Specifying the significance level
C. Bias analysis
D. Setting up the hypothesis
6. An analyst is testing if the mean profits of a sector are greater than $2m. A sample of 30 is
taken, and the value of the test statistic is 1.6. If you choose a 5% significance level you would:
A. Fail to reject the null hypothesis and conclude that the population mean is less than or equal
to $2m
B. Reject the null hypothesis and conclude that the population mean is equal to $2m
C. Fail to reject the null hypothesis and conclude that the population mean is greater than $2m
D. Reject the null hypothesis and conclude that the population mean is equal to zero
7. In hypothesis testing, which of the below best defines a Type I error?

A. Not rejecting a true null hypothesis
B. Rejecting a false null hypothesis
C. Rejecting a true null hypothesis
D. Not ejecting a false null hypothesis

SOLUTIONS
1. D
We are testing the coefficients of the two independent variables i.e. b1 and b2. Because we are
interested in whether there is a statistically significant relationship or not, our hypotheses will
be:
H o : b1 ≥ 0
Ha: b1 < 0
Recall also that Ho is an equality statement, whereas Ha is not. Ho states that there is no
relationship between the variables, whereas Ha states that there is.
Thus, a rejection of Ho would indicate a statistically significant relationship between the
variables, whereas a non-rejection of Ho would indicate that there is no statistically significant
relationship between the variables.
2. C
A two-tailed test always uses equal/not equal to with the population mean used rather than
the sample mean.
3. A
The hypothesis test performed is a one-tailed test. The two samples are less than 30, which
means that the differences in means will be t-distributed with n1 + n2 - 2 degrees of freedom,
i.e. df = 24. For an alpha (significance) of 0.05 the t-tables, for 24 df give us a critical value of
1.710882 = 1.7109.
4. A
The test statistic would be calculated as follows: (Observed value - Hypothesized value)
/Standard error (13 - 12)/(5/SQRT 30) = 1.09545 Note: Standard error = Standard
deviation/SQRT(n).
5. C
The steps in hypothesis testing are: 1) Stating the hypotheses 2) Identifying the appropriate test
statistic and its probability distribution 3) Specifying the significance level 4) Stating the
decision rule 5) Collecting the data and calculating the test statistic 6) Making the statistical
decision 7) Making the economic or investment decision.
6. A
The analyst is trying to test if the mean profits are greater than $2m; this is the alternative
hypothesis as by convention we set up a null as being that which we are trying to disprove. The
null hypothesis is therefore that the mean profits are less than or equal to $2m. The critical Z
value for a 1 tailed test at the 5 percent significance level is 1.65. Since the calculated value of 1.6
is less than 1.65, we fail to reject the null hypothesis.
7. C
C is the definition of a Type I error. A and B are accurate decisions

Topic - 18
Linear Regression [QA - 7]
Learning Objectives
Linear Regression
LO 18 a: Describe the models which can be estimated using linear regression and differentiate them
from those which cannot.
LO 18 b: Interpret the results of an OLS regression with a single explanatory variable.
LO 18 c: Describe the key assumptions of OLS parameter estimation.
LO 18 d: Characterize the properties of OLS estimators and their sampling distributions.
LO 18 e: Construct, apply, and interpret hypothesis tests and confidence intervals for a single
regression coefficient in a regression.
LO 18 f: Explain the steps needed to perform a hypothesis test in a linear regression.
LO 18 g: Describe the relationship between a t-statistic, it’s p-value, and a confidence interval.

QUESTIONS
1. Which of the following statements regarding regression analysis are correct?

I. The independent variable is generally labeled X.
II. The dependent variable is generally labeled Y.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Which of the following statements regarding the error term in the regression equation is
correct? The equation for the error term is denoted as follows:
3. Which of the following statements regarding the residuals or error term is not correct?
A. A residual is simply the vertical distance between a point on the scatter plot and the line of
regression.
B. The better the fit of the line to the points, the smaller the residuals.
C. Should the line of regression pass through a particular point, the residual value of that
point will be 1.
D. The reason for squaring the residuals is to eliminate the cancellation effect of big positive
and big negative residuals.
4. Which of the following statements regarding regression analysis are correct?

I. In the event that there was not a linear relationship between the dependent variable and
the parameters we could not use a linear regression model
II. In the event that there was not a linear relationship between the variables themselves we
could not use a linear regression model.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
E.
5. Consider the following regression equation:
ŷ  11.581 + 6.917x.
Assume that the value of the independent variable is equal to 10.5. What is the value of the
dependent variable?
A. 0
B. 11.581
C. 3.192

D. 84.209
6. Which of the following assumption is not a key assumption of OLS for estimation of
parameters?
A. All observations (X and Y) are independent and identically distributed (iid).
B. Large outliers are not expected to be present in the data as this is likely to distort the
regression results.
C. The expected value of the error (residual) term is 0.
D. There is a linear (straight line) relationship between the dependent variable Y and
the independent variable X.
7. Which of the following statements regarding the benefits of using OLS estimators are correct?
I. Due to its popularity and wide use, several software packages are available and are widely
used.
II. The regression coefficients used in OLS are easily understood and represent the desirable
properties of an estimator.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
8. Which of the following statements regarding the properties of OLS estimators are correct?
I. OLS estimators have got their own sampling distributions which allow us to estimate the
population parameters, such as the population mean, intercept and slope.
II. Assuming that b0 and b1 are normally distributed we can then assume that the sampling
distributions of b0 and b1 are unbiased and consistent estimators of the population
parameters B0 and B1.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.

SOLUTIONS
1. C
The independent variable is generally labeled X and the dependent variable Y. Common sense
will tell you which is which.
2. B
The equation for the error term is denoted as follows:
i =Yi – (b0 + b1Xi)
3. C
Before moving on, let us explain a little about residuals or error term.
A residual is simply the vertical distance between a point on the scatter plot and the line of
regression. The regression line that is used is the one that minimizes the sum of the squares of
the residuals. In this way, it is the best line to use, and hence the name ‘least squares line of
regression’.
Obviously, the better the fit of the line to the points, the smaller will be the distances between
the line and the points, and hence the smaller the residuals. Should the line of regression pass
through a particular point, the residual value of that point will be 0.
The reason for squaring the residuals is to eliminate the cancellation effect of big positive and
big negative residuals. By squaring first, all residual distances become positive, so the best line
can then be found.
4. A
In the event that there was not a linear relationship between the dependent variable and the
parameters we could not use a linear regression model.
In the event that there was not a linear relationship between the variables themselves we could
still use a linear regression model.
5. D
ŷ  11.581 + 6.917x.
Substitute the 10.5 into the equation.
6. D
 All observations (X and Y) are independent and identically distributed (iid).
 Large outliers are not expected to be present in the data as this is likely to distort
the regression results.
 The expected value of the error (residual) term is 0. Another way of saying this is that E[i]
= 0. This assumption states that, when the regression line is ‘best’, the positive residuals
will exactly cancel out the negative residuals, making the average value of the error term 0.

Additional assumptions:
There is a linear (straight line) relationship between the dependent variable Y and the
independent variable X.
7. B
Due to its popularity and wide use, several software packages are available and are
widely used.
The regression coefficients used in OLS are easily understood and represent the desirable
properties of an estimator, in other words, they unbiased, consistent and efficient.
8. C
OLS estimators have got their own sampling distributions which allow us to estimate the
population parameters, such as the population mean, intercept and slope.
Assuming that b0 and b1 are normally distributed we can then assume that the sampling
distributions of b0 and b1 are unbiased and consistent estimators of the population parameters
B0 and B1.

Topic - 19
Regression with Multiple Explanatory Variables [QA - 8]
Learning Objectives
Regression with Multiple Explanatory Variables
LO 19 a: Distinguish between the relative assumptions of single and multiple regression.

LO 19 b: Interpret regression coefficients in a multiple regression.
LO 19 c: Interpret goodness of fit measures for single and multiple regressions, including R2 and
adjusted- R2.
LO 19 d: Construct, apply, and interpret joint hypothesis tests and confidence intervals for multiple
coefficients in a regression.

QUESTIONS
1. When looking at the relationship between the percentage bid-ask spread (as measured by the
spread divided by the stock price) and the number of market makers in the market and
the company’s market capitalization.
The following variables are applicable:
Percentage bid-ask spread = dependent variable = Y.
Number of market makers = independent variable 1 = X1.
Company’s market capitalization = independent variable 2 = X2.
The regression is estimated using data from 1819 NASDAQ-listed stocks from December 2002.
The following multiple regression equation is formulated:
Yi = b0 + b1X1i + b2X2i + i
Where:
Yi = Percentage bid-ask spread
b0 = Intercept
X1i = Number of market makers
X2i = Market capitalization of the company
The results from regressing Y on X1 and X2 are presented in the table below:
Which of the following most accurately represents the multiple regression equation?
A. Yi = -0.1369 + 0.0673(MM) + 0.0246 (Cap) + i
B. Yi = -0.7586 + -0.2790(MM) + -0.6635 (Cap) + i
C. Yi = -5.5416 + -0.4.1427(MM) + -27.0087 (Cap) + i
D. Yi = -0.7586 + 0.1369(MM) + -5.5416 (Cap) + i
spread divided by the stock price) and the number of market makers in the market and the
company’s market capitalization.
Company’s market capitalization = independent variable 2 = X2.

Yi = b0 + b1X1i + b2X2i + i
Where:
b0 = Intercept
Assuming the following regression equation:

Yi = -0.7586 + -0.2790(MM) + -0.6635 (Cap) + i
What would the effect on the bid-ask spread be if we added one more market makers?
A. It would increase by 0.2790
B. It would decrease by 0.2790
C. It would increase by 0.6635
D. It would decrease by 0.6635
3. Which of the following statements regarding multiple regression analysis are correct?
I. The general equation for the multiple linear regression model is: Yi = b0 + b1X1i + b2X2i
+………+ bkXki+ i
II. The residual error can be expressed as follows: î  Yi  Yî
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
4. When dealing with the R2 (coefficient of determination), which of the following equations is not
correct?
A. R2 = Total variation (TSS) – unexplained variation (SSR) / Total variation (TSS)
B. R2 = Explained variation / Total variation
C. R2 = ESS / TSS
D. R2 = RSS / TSS
5. When dealing with adjusted R2, which of the following is not a reason that adjusted R2 was
introduced as a measure?
A. One of the issues with the Coefficient of determination (R2) is that it itself may not a reliable
measure of the explanatory power of the additional variables.

B. Almost always, as additional independent variables are added to the model the R2 will
increase even if the contribution of the new variables is not statistically significant.
C. A high R2 may be as a result of a large number of independent variables in the model and
not as a result of their explanatory power.
D. Almost always, as additional independent variables are added to the model the R2 will
increase unless the contribution of the new variables is not statistically significant.
6. Which of the following assumptions is not an assumption of the multiple linear regression
model?
A. The expected value of the error (residual) term is 0.
B. The variance of the error term is the same for all observations.
C. The residuals are independently distributed.
D. The error term is lognormally distributed.
7. When assessing the effects of multicollinearity on our analysis, which of the following
statements are not correct?
A. The slope coefficients remain consistent. However, they tend to be unreliable.
B. The standard errors tend to be inflated and as such the test-statistics are not powerful and
lack the ability to reject the null hypothesis.
C. This leads to type I errors.
D. This leads to type II errors.
8. Which of the following statements regarding the hypothesis test are not correct?
A. Ho: b1 ≥0
B. Ha: b1 < 0
C. Ho states that there is no relationship between the variables
D. A failure to reject Ho would indicate a statistically significant relationship between the
variables.
spread divided by the stock price) and the number of market makers in the market and the
company’s market capitalization.
Yi = b0 + b1X1i + b2X2i + i
Where:
b0 = Intercept

Assuming a confidence level of 99%, what is the confidence interval for X2i (Market
capitalization of the company)?
A. -0.221; -0.337
B. -0.221; -0.0246
C. -0.026; -0.337
D. 0.221; 0.337
10. Which of the following statements regarding the F-test are correct?
I. The F-test assesses how well the independent variables explain the variation in the
dependent variable.
II. The alternative hypothesis relating to the F-test will be: Ho: b1 = b2 = b3 = … = bk = 0
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
11. Which of the following statements are not correct when stating the formula for the F-stat?
A. F = Mean regression sum of the squares / Mean squared error
B. F = MSR / MSE
C. F = (ESS/k) / (SSR/n-k-1)
D. F = (TSS/k) / (SSR/n-k-1)
12. The following ANOVA table has been presented:
Based on the ANOVA table, the F-stat is:

A. 1088.83
B. 1340.82
C. 1819.00
D. 4917.93

SOLUTIONS
1. B
Yi = b0 + b1X1i + b2X2i + i
Keep in mind that we use the regression coefficient column for our regression coefficients.
Yi = -0.7586 + -0.2790(MM) + -0.6635 (Cap) + i
2. B
Based on our example, assuming that we increase the number of market makers by one unit,
we would expect the percentage bid-ask spread divided by the stock price to decrease
by 0.2790. This is assuming that we hold the other independent variable constant.
3. B
The general equation for the multiple linear regression model is:
Yi = b0 + b1X1i + b2X2i +………+ bkXki+ i
The residual can be expressed as follows:
î  Yi  Yî
4. D
R2 = Total variation (TSS) – unexplained variation (SSR) / Total variation (TSS)
Which is also equal to:
R2 = Explained variation / Total variation
R2 = ESS / TSS
5. D
Adjusted R2
One of the issues with the Coefficient of determination (R2) is that it itself may not a reliable
measure of the explanatory power of the additional variables. Almost always, as additional
independent variables are added to the model the R2 will increase. This is so even if the
contribution of the new variables is not statistically significant. Based on this a high R2 may be
as a result of a large number of independent variables in the model and not as a result of their
explanatory power. This is referred to as overestimating the regression.
6. D
 The variance of the error term is the same for all observations. The statistical term for this is
homoskedasticity.
 The residuals are independently distributed. Here, we are saying that the residual for one
observation is not correlated with that of another observation. We say that there is no
autocorrelation among the residual terms.

 The error term is normally distributed. This assumption makes the mathematics easier to
work with.
7. C
What effect does multicollinearity have on our analysis?
The slope coefficients remain consistent. However, they tend to be unreliable.
The standard errors tend to be inflated and as such the test-statistics are not powerful and lack
the ability to reject the null hypothesis.
This leads to type II errors – that we conclude that the variable is not statistically significant.
8. D
We are testing the coefficients of the two independent variables i.e. b1 and b2. Because we are
interested in whether there is a statistically significant relationship or not, our hypotheses will
be:
H o : b1 ≥ 0
Ha: b1 < 0
Recall also that Ho is an equality statement, whereas Ha is not. Ho states that there is no
relationship between the variables, whereas Ha states that there is.
Thus, a rejection of Ho would indicate a statistically significant relationship between the
variables, whereas a non-rejection of Ho would indicate that there is no statistically significant
relationship between the variables.
9. A
In our case degrees of freedom will be:
= 1819 – 2 (number of independent variables) – 1 (which is the slope)
= 1816
Critical values:
Look up on the t-test tables at 1816 degrees of freedom. However, the table does not go up to
that high so we will use degrees of freedom of 200. The critical value is 2.345.
Putting this all together, we get a 99% confidence interval to be:
j 
bˆ  t  sbˆ
c j 
= -0.6635  2.345 * 0.0246
= -0.2790  0.058
= (-0.221 ; -0.337)
10. A
The F-test assesses how well the independent variables explain the variation in the dependent
variable.
Based on this the null hypothesis will be:
H o : b1 = b 2 = b 3 = … = b k = 0

11. D
Based on the above four variables, the formula for the F-stat is:
F = Mean regression sum of the squares / Mean squared error
F = MSR / MSE
F = (ESS/k) / (SSR/n-k-1)
12. A
F = (ESS/k) / (SSR/n-k-1)
F = (2681.6482 / 2) / (2236.2820/ 1816)
F = 1088.8325

Topic - 20
Regression Diagnostics [QA - 9]
Learning Objectives
Regression Diagnostics
LO 20 a: Explain how to test whether a regression is affected by heteroskedasticity.

LO 20 b: Describe approaches to using heteroskedastic data.
LO 20 c: Characterize multicollinearity and its consequences; distinguish between multicollinearity
and perfect collinearity.
LO 20 d: Describe the consequences of excluding a relevant explanatory variable from a model and
contrast those with the consequences of including an irrelevant regressor.
LO 20 e: Explain two model selection procedures and how these relate to the bias-variance trade-off.
LO 20 f: Describe the various methods of visualizing residuals and their relative strengths.
LO 20 g: Describe methods for identifying outliers and their impact.
LO 20 h: Determine the conditions under which OLS is the best linear unbiased estimator.

QUESTIONS
1. Which of the following statements regarding when the omitted variable bias occurs are correct?
I. There is a correlation between the omitted variable and the movement of the independent
variable.
II. The omitted variable assists in the determination of the dependent variable.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
2. Which of the following statements regarding homoscedasticity and heteroskedasticity are not
correct?
A. Homoskedasticity is when the variance of the error term is the same for all observations.
B. Heteroskedasticity is when the variance of the error term is not the same for all
observations.
C. Heteroskedasticity can be broken down into conditional and unconditional
heteroskedasticity.
D. Unconditional heteroskedasticity presents an issue for the regression analysis.
3. Which of the following statements regarding multiple regression analysis are correct?
I. The general equation for the multiple linear regression model is: Yi = b0 + b1X1i + b2X2i
+………+ bkXki+ i

II. The residual error can be expressed as follows: 
ˆ i  Yi  Yi
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
4. Which of the following assumptions is not an assumption of the multiple linear regression
model?
A. The expected value of the error (residual) term is 0.
B. The variance of the error term is the same for all observations.
C. The residuals are independently distributed.
D. The error term is lognormally distributed.
5. When assessing the effects of multicollinearity on our analysis, which of the following
statements are not correct?
A. The slope coefficients remain consistent. However, they tend to be unreliable.
B. The standard errors tend to be inflated and as such the test-statistics are not powerful and
lack the ability to reject the null hypothesis.
C. This leads to type I errors.
D. This leads to type II errors.

6. When dealing with multicollinearity, which of the following statements are correct?
I. One of the ways to detect multicollinearity is a high R2 and a statistically significant F-
statistic but the t-tests indicate that none of the individual coefficients is significantly
different to zero.
II. The best way to correct for multicollinearity is to leave out one or more of the
variables that are correlated.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.

SOLUTIONS
1. C
Omitted variable bias occurs when the following two conditions are met:
 There is a correlation between the omitted variable and the movement of the independent
variable.
 The omitted variable assists in the determination of the dependent variable.
2. D
The variance of the error term is the same for all observations. This is called homoskedasticity.
Heteroskedasticity is when the variance of the error term is not the same for all observations.
Heteroskedasticity can be broken down into conditional and unconditional heteroskedasticity.
Unconditional heteroskedasticity
This occurs when the heteroskedasticity is unrelated to level of the independent variables. In
other words it does not move around as the independent variables are changed.
This violates the regression assumptions but does not present an issue for the regression
analysis.
3. B
The general equation for the multiple linear regression model is:
Yi = b0 + b1X1i + b2X2i +………+ bkXki+ i
The residual can be expressed as follows:


ˆ i  Yi  Yi
4. D
 The variance of the error term is the same for all observations. The statistical term for this is
homoskedasticity.
 The residuals are independently distributed. Here, we are saying that the residual for one
observation is not correlated with that of another observation. We say that there is no
autocorrelation among the residual terms.
 The error term is normally distributed. This assumption makes the mathematics easier to
work with.
5. C
What effect does multicollinearity have on our analysis?
The slope coefficients remain consistent. However, they tend to be unreliable.
The standard errors tend to be inflated and as such the test-statistics are not powerful and lack
the ability to reject the null hypothesis.
This leads to type II errors – that we conclude that the variable is not statistically significant.

6. C
How are we able to detect multicollinearity?
One of the ways to detect multicollinearity is a high R2 and a statistically significant F-statistic
but the t-tests indicate that none of the individual coefficients is significantly different to zero.
How do we correct multicollinearity?

The best way to correct for multicollinearity is to leave out one or more of the variables that are
correlated.

Topic - 21
Stationary Time Series [QA-10]
Learning Objectives
Stationary Time Series
LO 21 a: Describe the requirements for a series to be covariance stationary.

LO 21 b: Define the autocovariance function and the autocorrelation function.
LO 21 c: Define white noise, describe independent white noise and normal (Gaussian) white noise.
LO 21 d: Define and describe the properties of autoregressive (AR) processes.
LO 21 e: Define and describe the properties of moving average (MA) processes.
LO 21 f: Explain how a lag operator works.
LO 21 g: Explain mean reversion and calculate a mean-reverting level.
LO 21 h: Define and describe the properties of autoregressive moving average (ARMA) processes.
LO 21 i: Describe the application of AR, MA, and ARMA processes.
LO 21 j: Describe sample autocorrelation and partial autocorrelation.
LO 21 k: Describe the Box-Pierce Q-statistic and the Ljung-Box Q statistic.
LO 21 l: Explain how forecasts are generated from ARMA models.
LO 21 m: Describe the role of mean reversion in long-horizon forecasts.
LO 21 n: Explain how seasonality is modeled in a covariance-stationary ARMA.

QUESTIONS
1. In order for the assumption of covariance stationary to be valid we need several conditions to
be in place. Which of the following assumptions are not correct?
A. The expected value of the time-series must be constant and finite for all periods.
B. The variance of the time-series must be constant and finite for all periods.
C. The covariance of the time-series with itself over a fixed number of periods (both past and
future) must be both constant and finite.
D. The covariance must resemble a white noise process.
2. Which of the following statements regarding the autocorrelation function are correct?
I. The autocorrelation function is when the covariances have been converted to correlations.
II. The correlation will sit between -1 and +1.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
3. Archie Bal, FRM, is busy looking at the implications of time series data that is not covariance
stationary. Which of the following statements in this regard are correct?
I. When time series data is not covariance stationary, the estimated regression will not
work unless we transform the data.
II. Transforming the data involves changing the data in such a way that we can find a model-
able covariance stationary property and work with this.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
4. Which of the following statements are correct regarding white noise?

A. White noise exhibits a mean of one and a constant variance.
B. White noise exhibits a mean of zero and a constant variance.
C. White noise exhibits a mean of one and a variance of two.
D. White noise and zero-mean white noise are different concepts.
5. When making use of the white noise process, which of the following statements are not correct?
A. The model’s forecast error should follow the white noise process.
B. The model’s forecast error should exhibit heteroskedasticity.
C. In the event that the autocorrelation function is perfectly serially uncorrelated then the
process would show zeros for all the lags / displacements.
D. The autocorrelation at τ = 0 will always equal one.

6. A distributed lag is when the model assigns weights to past values in the time series. Which of
the following equations regarding the model are correct?
I. Yt- + 0.4 yt-1 + 0.3 yt-2 + 0.2 yt-3
II. Yt-(1 + 0.4L+ 0.3L2 + 0.2L3)
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
7. Which of the following statements are not correct when dealing with the standard deviation of
its autocorrelations or partial correlations in a time series?
A. In a time series, the standard deviation of its autocorrelations or partial correlations,
1
assuming T observations, will equal .
T
B. One of the ways to determine if a time series is considered white noise is to display their
2
autocorrelation and partial autocorrelation functions in bands of 
T
C. If the time series is white noise then we could expect 95% of the sample autocorrelations
2
and sample partial autocorrelations to fall within the interval of 
T
D. An autoregression is a nonlinear regression of a time series against the past values of the
series.

SOLUTIONS
1. D
In order for the assumption of covariance stationary to be valid we need the following
conditions to be in place:
1. The expected value of the time-series must be constant and finite for all periods.
2. The variance of the time-series must be constant and finite for all periods.
3. The covariance of the time-series with itself over a fixed number of periods (both past and
future) must be both constant and finite.
2. C
Autocorrelation function – The autocorrelation function is when the covariances have been
converted to correlations. As we have seen in earlier topics, the correlation will sit between -1
and +1.
3. C
When time series data is not covariance stationary, the estimated regression will not work
unless we transform the data. Transforming the data involves changing the data in such a way
that we can find a model-able covariance stationary property and work with this.
4. B
A time series that exhibits a zero mean and a constant variance is called white noise or zero-
mean white noise. When we say zero mean we are talking about it’s unconditional mean.
5. B
When using the white noise process, the model’s forecast error should follow the white noise
process. If not, then the errors can be forecast based on past values. However, the model itself is
inaccurate and cannot be used to predict accurately. In the event that the autocorrelation
function is perfectly serially uncorrelated then the process would show zeros for all the lags /
displacements. Keep in mind that the autocorrelation at τ = 0 will always equal one. The reason
for this is that this is the correlation of the series with itself.
6. C
A distributed lag is when the model assigns weights to past values in the time series. The
model will look as follows:
Yt- + 0.4 yt-1 + 0.3 yt-2 + 0.2 yt-3
This can be re-written as follows:
Yt-(1 + 0.4L+ 0.3L2+ 0.2L3)
7. D
In a time series, the standard deviation of its autocorrelations or partial correlations, assuming
1
T observations, will equal
T

One of the ways to determine if a time series is considered white noise is to display their
2
autocorrelation and partial autocorrelation functions in bands of 
T
If the time series is white noise then we could expect 95% of the sample autocorrelations and
sample partial autocorrelations to fall within this interval.

Topic - 22
Non-Stationary Time Series [QA - 11]
Learning Objectives
Non-Stationary Time Series
LO 22 a: Describe linear and nonlinear time trends.

LO 22 b: Explain how to use regression analysis to model seasonality.
LO 22 c: Describe a random walk and a unit root.
LO 22 d: Explain the challenges of modeling time series containing unit roots.
LO 22 e: Describe how to test if a time series contains a unit root.
LO 22 f: Explain how to construct an h-step-ahead point forecast for a time series with seasonality.
LO 22 g: Calculate the estimated trend value and form an interval forecast for a time series.

QUESTIONS
1. Which of the following statements regarding linear models are not correct?
A. A linear trend is a trend that can be graphed using a straight line.
B. In this model the dependent variable (Y) changes at rate constant with time.
C. The equation for the linear trend model is: Yt = b0 + b1t+ i
D. The equation for the linear trend model is: Yt = b0 + b1(t) + b2(t)2 + i
2. Which of the following statements regarding the quadratic model are correct?
I. The quadratic model can model trends by working with the sign and the level of the
coefficients.
II. When dealing with time-series data, the data will often exhibit growth factors that grow at
a continuous rate
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
3. David Jones, FRM, is in the process of constructing a quadratic model, he needs assistance with
his model. Assist David by informing him which of the following signs of the variables will
result in a U shape when graphed?
A. Both b1 and b2 are negative.
B. Both b1 and b2 are positive.
C. When b1 is positive and b2 is negative.
D. When b1 is negative and b2 is positive.
4. An analyst wishes to predict future inflation rates.

Monthly U.S. CPI information is used for this purpose. The data runs from January 1995 for a
period of 228 months.

The model for this will be: Y  bˆ  bˆ t
t o 1
The table below is an extract of the regression results:
The estimated regression equation is:


Yt  2.8853  0.0038t

Calculate the inflation rate for month 120.

A. 1.9733%
B. 2.8853%
C. 2.4290%
D. 0.0038%
5. Which of the following statements are correct when looking at a time series model?
A. If the variable grows at a constant rate = use a linear model.
B. If the variable grows at a constant amount = use a log-linear model.
C. In the event that the plotted points resemble a curved shape, this would indicate a
linear line and a linear model would be more appropriate.
D. In the event that the plotted points resemble a curved shape, this is an indication that the
error terms are serially correlated (autocorrelation).
6. Which of the following regarding random walks is correct?

A. A random walk is a time series in which the value of the series in one period is the value of
the series in the previous period plus an unpredictable random error.
B. The random walk equation is a special case of an AR(1) model with b0 =1 and b1 = 0.
C. The expected value of εt is 1.
D. The random walk equation is a special case of an AR(2) model with b0 =1 and b1 = 0.
7. Which of the following regarding random walks is incorrect?

A. The best forecast of xt that can be made in period t − 1 is xt−1.
B. Because exchange rates follow a random walk, the best forecast of the future exchange rate
is the current exchange rate.
C. A random walk has a defined mean-reverting level.
D. A random walk is a time series in which the value of the series in one period is the value of
the series in the previous period plus an unpredictable random error.
8. Which of the following regarding the Dickey–Fuller test for a unit root is incorrect?
A. The regression-based unit root test is based on a transformed version of the AR(1) model xt
= b0 + b1xt−1 + εt, by subtracting xt−1 from both sides of the AR(1) model producing b0 + g1xt–1
+ εt, E(εt) = 0 where g1 = (b1 − 1).
B. If b1 = 1, then g1 = 0 and thus a test of g1 = 0 is a test of b1 = 1.
C. To conduct the test, one calculates a t-statistic in the conventional manner for g^1 and using
conventional critical value tables.
D. Dickey and Fuller convert the AR(1) model and run a simple regression.

SOLUTIONS
1. D
Linear Trend Model
A linear trend is a trend that can be graphed using a straight line. In this model the dependent
variable (Y) changes at rate constant with time.
The equation for the linear trend model is:
Yt = b0 + b1t+ i
2. A
The quadratic model can model trends by working with the sign and the level of the
coefficients.
When dealing with time-series data, the exponential model states that the data will often
exhibit growth factors that grow at a continuous rate, i.e. exponential growth.
3. D
The quadratic model can model trends by working with the sign and the level of the
coefficients.
By way of example: Assuming that both b1 and b2 are negative, the trend will increase at a
decreasing rate. Assuming that both b1 and b2 are positive, the trend will increase at an
increasing rate.
When b1 is positive and b2 is negative, the trend will exhibit an upside down U shape.
When b1 is negative and b2 is positive, the trend will exhibit a U shape.
4. C
Assuming that we wanted to predict the inflation rate at month 120, we would do so as follows:
ŷ120  2.8853  0.0038 120 
ŷ120  2.429 percent
5. D
The question is which model should be used with a particular time-series?
If the variable grows at a constant amount = use a linear model.
If the variable grows at a constant rate = use a log-linear model.
In the event that the plotted points resemble a curved shape, this would indicate a non-linear
line and a log-linear model would be more appropriate. This is an indication that the error
terms are serially correlated (autocorrelation) and therefore by using the natural log, the
regression line will fit the data better. This is often the case for financial data such as stock
prices.
6. A
The random walk equation is a special case of an AR(1) model with b0 = 0 and b1 = 1. The
expected value of εt is 0.

7. C
A random walk has an undefined mean-reverting level of 0/0.
8. C
To conduct the test, one calculates a t-statistic in the conventional manner for g^1 but using a
revised set of critical values computed by Dickey and Fuller (larger in absolute value than the
conventional values).

Topic - 23
Measuring Return, Volatility, and Correlation [QA - 12]
Learning Objectives
Measuring Return, Volatility, and Correlation
LO 23 a: Calculate, distinguish, and convert between simple and continuously compounded

returns.
LO 23 b: Define and distinguish between volatility, variance rate, and implied volatility.
LO 23 c: Describe how the first two moments may be insufficient to describe non-normal
distributions.
LO 23 d: Explain how the Jarque-Bera test is used to determine whether returns are normally
distributed.
LO 23 e: Describe the power law and its use for non-normal distributions.
LO 23 f: Define correlation and covariance and differentiate between correlation and dependence.
LO 23 g: Describe properties of correlations between normally distributed variables when using a
one-factor model.

QUESTIONS
1. Which of the following statements regarding implied volatility are correct?

A. To calculate implied volatility we use the standard deviation of returns over a one-year
period.
B. To calculate implied volatility we use the standard deviation of returns over a one-day
period.
C. To calculate implied volatility we use the square of the volatility measure (standard
deviation).
D. To calculate implied volatility we set the Black-Scholes-Merton (BSM) model equal to the
market price of the option – we can work backwards to infer the volatility.
2. Which of the following statements regarding the power law are correct?
I. The power law indicates that the probability of extreme events is low but are still higher
than indicated by the normal distribution.
II. The power law can be expressed as follows: P(V > X) = K x X-α
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
3. Which of the following statements regarding covariance is not correct?

A. The covariance of returns is 0 if the returns on the two assets are unrelated.
B. The covariance of returns is positive if the return on both assets is above (or below) the
corresponding expected value at the same time.
C. The covariance of returns is negative if, when the return on one asset is above its expected
value, the return on the other asset is below its expected value and vice versa.
D. The covariance of returns ranges between -1 and +1.
4. Which of the following statements regarding the single factor model are correct?
I. A commonly used single factor model is the capital asset pricing model (CAPM).
2
II. The following equation will be used for single factor model: Ui  iF  1   Z i
i
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
5. Which of the following statements regarding the normal distribution are not correct?
A. The area under the graph equals 1.
B. The graph is perfectly symmetrical and therefore has zero skewness.
C. The tails get smaller and smaller the further we move away from the mean.
D. Empirically, the normal distribution does not approximate real-world data very well.

6. Which of the following statements regarding kurtosis is not correct?

A. A peaked distribution is referred to as leptokurtic.
B. A flat or non-peaked distribution is referred to as platykurtic.
C. The middle-of-the-road distribution is referred to as mesokurtic.
D. Excess kurtosis is equal to 3.
7. Which of the following steps is not part of the Spearman rank correlation process?
A. Rank the observations from lowest to highest.
B. Allocate the ranks to the relevant data
C. Calculate the difference in the rankings of each observation and square them.
D. Apply the Spearman rank correlation equation.
8. Which of the following portfolio standard deviation calculations is correct for a portfolio with
zero correlation?
9. Which of the following is not a cause of non-normality?

A. Autocorrelation.
B. Illiquidity.
C. Nonlinearity.
D. Heteroskedasticity

SOLUTIONS
1. D
To calculate implied volatility we set the Black-Scholes-Merton (BSM) model equal to the
market price of the option – we can work backwards to infer the volatility.
2. C
The issue that we experience in real life situations is that the distributions are seldom normal
distributions and often exhibit fatty tails with much larger standard deviations. One method of
dealing with this is to apply the power law to the normal distribution. The power law indicates
that the probability of extreme events is low but are still higher than indicated by the normal
distribution.
The power law can be expressed as follows:
P(V > X) = K x X-α
3. D
 The covariance of returns is 0 if the returns on the two assets are unrelated.
 The covariance of returns is positive if the return on both assets is above (or below) the
corresponding expected value at the same time.
 The covariance of returns is negative if, when the return on one asset is above its expected
value, the return on the other asset is below its expected value and vice versa.
 The correlation coefficient ranges between -1 and +1.
4. C
A commonly used single factor model is the capital asset pricing model (CAPM). The CAPM
has a return component for each asset. This return has a systematic component (Beta) that is
correlated with the market portfolio return and an unsystematic component that is independent
of the other stocks on the market.
The following equation will be used:
2
Ui  iF  1   Z i
i
5. D
Also, because the graph is symmetrical about the mean, and because the area under the graph
equals 1 (as it is a probability), it follows that each side of the graph represents 50% of the total
area.
The graph is perfectly symmetrical and therefore has zero skewness.
Its’ kurtosis is equal to 3. You will recall that excess kurtosis is measured relative to the number
3. We can, therefore, say that the graph has zero excess kurtosis.
The tails get smaller and smaller the further we move away from the mean. However, they
never equal zero – they extend indefinitely. These are called asymptotic tails. Empirically, the
normal distribution approximates real-world data very well.

6. D
A peaked distribution is referred to as leptokurtic. This type of distribution has most of its
values clustered closely around the mean.
A flat or non-peaked distribution is referred to as platykurtic. This type of distribution is very
flat and has large dispersion statistics.
The middle-of-the-road distribution is referred to as mesokurtic. A normal distribution fits this
description.
Excess kurtosis – Kurtosis minus 3. This is a measure used so as to make the excess kurtosis of
a normal distribution equal to 0.
7. A
1. Rank the observations from highest to lowest.
2. Allocate the ranks to the relevant data
3. Calculate the difference in the rankings of each observation and square them.
4. Apply the Spearman rank correlation equation.
8. B
When there is no correlation between the two assets, then the third term in the equation
disappears (since the correlation is zero). Based on this, the portfolio standard deviation
calculation is as follows:
1
 2 2 2 2 2
port  w   w  
 1 i 2 2
9. D
1. Autocorrelation. We have discussed this in the previous section. Autocorrelation causes
outcomes that are more extreme than those predicted by the normal distribution.
2. Illiquidity. As a result of illiquidity, which is prevalent in alternative investments, values
are often estimated as opposed to using market values. This results in positive
autocorrelation, which produces non-normal returns.
3. Nonlinearity. An example of non-linear returns is that of a call option that exhibits non-
symmetrical returns.

Topic - 24
Simulation and Bootstrapping [QA - 13]
Learning Objectives
Simulation and Bootstrapping
LO 24 a: Describe the basic steps to conduct a Monte Carlo simulation.

LO 24 b: Describe ways to reduce Monte Carlo sampling error.
LO 24 c: Explain the use of antithetic and control variates in reducing Monte Carlo sampling error.
LO 24 d: Describe the bootstrapping method and its advantage over Monte Carlo simulation.
LO 24 e: Describe pseudo-random number generation and how a good simulation design alleviates
the effects the choice of the seed has on the properties of the generated series.
LO 24 f: Describe situations where the bootstrapping method is ineffective.
LO 24 g: Describe the disadvantages of the simulation approach to financial problem solving.

QUESTIONS
1. Which of the following statements is not an application of Monte Carlo simulation?

A. Assessing and estimating the potential risks of a particular trading strategy
B. Studying returns on assets that have complicated features associated with them, or that
have parameters that change constantly
C. Assessing and estimating the success rate of a particular trading strategy
D. Examining the interaction between different classes of investments using very small
samples.
2. Consider the following information:
What is the standard error if we increase the number of simulations to 500?

A. 18
B. 9.04
C. 8.04
D. 100
3. Which of the following statements regarding the antithetic variates are correct?
I. By using the antithetic variates we make use of a negative covariance between the original
sample and its compliment.
II. Using an antithetic variate approach the MCS sampling error is smaller smaller using this
approach than a standard MCS approach.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
4. Which of the following statements are not correct when dealing with the control variate
technique?
A. Another alternative that can be used to reduce the sampling error is called the control
variate technique.
B. The control variate technique removes a variable used (x) that has unknown properties and
replaces it with a variable (y) that has known properties.
C. Our new x* estimate will have a higher sampling error than our original x variable if the
control statistic and the statistic of interest exhibit a high degree of correlation.
D. The known variable (y) and the new MCS results (x*) will now have similar properties.

5. Which of the following statements are not correct when dealing with the Dickey-Fuller test?
A. In the Dickey-Fuller test, we can reuse the same sample set for each simulation run
while we test for different parameters of the Dickey-Fuller test.
B. This Dickey-Fuller test will serve to reduce the sampling variation among the different
experiments.
C. This Dickey-Fuller test will result in reduced sampling error but not increased accuracy.
D. By using the same data we reduce the variability of the sample error.
6. The_______can be used to generate random numbers randomly from historical data using the
replacement method.
A. The bootstrapping method
B. The pseudo-random number generating method
C. The Dickey-Fuller method
D. The Derivative method
7. Which of the following statements are not correct when dealing with the bootstrapping
method?
A. A disadvantage of using this method is that we are not making any assumptions regarding
the real distribution of the parameter estimate.
B. Using the real data will allow the inclusion of fatter tails – which in truth is more close to
the real world circumstances.
C. The bootstrapping method generates estimates that are closer to the real distribution
properties of the original data.
D. When using the bootstrapping method, the autocorrelations that were present in the
original data set will not be present in the sample.
8. Which of the following statements regarding a pseudo-random number generator are correct?
I. A commonly used pseudo-random number generator generates number sequences
between 0 and 1 with each number having an equal chance of being selected.
II. The numbers are drawn from a continuous distribution.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.
9. Which of the following statements regarding outliers and the bootstrapping method are
correct?
I. In the event of outliers, the estimates arrived at may be inaccurate depending on the
number of times the outliers are included in the sample.
II. An effective method of dealing with outsiders is to use a large sample size.
A. I only.
B. II only.
C. Both I and II.
D. Neither I or II.

10. Which of the following is not a disadvantage of the simulation approach to financial problem
solving?
A. The results may not be precise.
B. High costs of the calculations.
C. The results relate to the specific experiment only.
D. The results are easy to replicate.

SOLUTIONS
1. D
Applications of Monte Carlo simulation involve:
 Assessing and estimating the potential risks of a particular trading strategy
 Studying returns on assets that have complicated features associated with them, or that
have parameters that change constantly
 Assessing and estimating the success rate of a particular trading strategy
 Examining the interaction between different classes of investments using large sample
sizes.
2. C
3. C
By using the antithetic variates we make use of a negative covariance between the
original sample and its compliment.
As such the error terms are independent for two sets by way of the negative covariance term.
This results in the MCS sampling error always being smaller using this approach.
4. C
In the event that an experiment contains a wide range of possible outcomes – sampling error is
likely to be present. Another alternative that can be used to reduce the sampling error is called
the control variate technique. This technique removes a variable used (x) that has unknown
properties and replaces it with a variable (y) that has known properties.
Our new x* estimate will have a lower sampling error than our original x variable if the control
statistic and the statistic of interest exhibit a high degree of correlation. The known variable (y)
and the new MCS results (x*) will now have similar properties.
5. D
The Dickey-Fuller test.
If time series data is not covariance stationary then the estimated regression for AR(1) will not
work unless we transform the data. Based on this we need to test if a time series is covariance
stationary or not. We can use the Dickey Fuller test for this. In this test, we can reuse the same
sample set for each simulation run while we test for different parameters of the Dickey-Fuller

test. This will serve to reduce the sampling variation among the different experiments. This will
result in reduced sampling error but not increased accuracy.
6. A
The bootstrapping method can be used to generate random numbers randomly from historical
data using the replacement method.
7. A
An advantage of using this method is that we are not making any assumptions regarding the
real distribution of the parameter estimate. Rather, we are using the real data relating to this
parameter, which includes outliers and other type events. This will allow the inclusion of fatter
tails – which in truth is more close to the real world circumstances. Thus, this method generates
estimates that are closer to the real distribution properties of the original data. A possible
downside of this method is that autocorrelations that were present in the original data set will
not be present in the sample. The reason for this is that the variables are not drawn in the same
sequence as the original set of data.
8. A
A commonly used pseudo-random number generator generates number sequences between 0
and 1 with each number having an equal chance of being selected. The numbers can be drawn
from either a discreet or a continuous distribution.
9. C
In the event of outliers, the estimates arrived at may be inaccurate depending on the number of
times the outliers are included in the sample. The same outlier can be included several times as
replacement sampling is used. An effective method of dealing with outsiders is to use a large
sample size.
10. D
 The results may not be precise. This may occur when the model is not correctly specified.
 High costs of the calculations.
 The results relate to the specific experiment only.
 The results are difficult to replicate. This may occur when a seed is not chosen as the start
of the experiment.

B2_Additonal question bank

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

B2_Additonal question bank

Uploaded by

Copyright:

Available Formats

FRM PART – I – QUANTITATIVE ANALYSIS

LO 12 a: Describe an event and an event space.

Compiled by ulurn Page 76

1. Which of the following statements regarding properties of probabilities are correct?

Compiled by ulurn Page 77

6. Consider the following probability matrix:

9. Which of the following statements regarding Bayes’ formula are correct?

Compiled by ulurn Page 78

Compiled by ulurn Page 79

Compiled by ulurn Page 80

Note that 2 is included because of the equal sign, but 6 is not.

Compiled by ulurn Page 81

Compiled by ulurn Page 82

Compiled by ulurn Page 83

To do this for five years in a row =

Compiled by ulurn Page 84

LO 13 a: Describe and distinguish a probability mass function from a cumulative distribution

Compiled by ulurn Page 85

3. Consider the following probability distribution of Revenue for Tradealot:

The expected Revenue for Tradealot is:

4. Which of the following is not a property of expectations?

5. Which of the following statements regarding the PMF is not correct?

Compiled by ulurn Page 86

9. Consider the following data set: 0, 1, 1, 1, 997.

Compiled by ulurn Page 87

Compiled by ulurn Page 88

Compiled by ulurn Page 89

Common Univariate Random Variables

Compiled by ulurn Page 90

Compiled by ulurn Page 91

D. The normal distribution is asymmetrical about its mean.

8. Which of the following statements regarding a mixture distribution are correct?

Compiled by ulurn Page 92

Compiled by ulurn Page 93

The key properties of a lognormal distribution:

Compiled by ulurn Page 94

Multivariate Random Variables

Compiled by ulurn Page 95

1. Consider the following data:

Compiled by ulurn Page 96

Are the profits of Big firm and Small firm independent?

Compiled by ulurn Page 97

Compiled by ulurn Page 98

Compiled by ulurn Page 99

2. Consider the following data:

What is the median of the above data observations?

Compiled by ulurn Page 100

Compiled by ulurn Page 101

LO 17 a: Construct an appropriate null hypothesis and alternative hypothesis and distinguish

Compiled by ulurn Page 102

Compiled by ulurn Page 103

5. Which of the below does not represent a step of hypothesis testing?

7. In hypothesis testing, which of the below best defines a Type I error?

Compiled by ulurn Page 104

Compiled by ulurn Page 105

Compiled by ulurn Page 106

1. Which of the following statements regarding regression analysis are correct?

4. Which of the following statements regarding regression analysis are correct?

Compiled by ulurn Page 107

Compiled by ulurn Page 108

Compiled by ulurn Page 109

Compiled by ulurn Page 110

Regression with Multiple Explanatory Variables

LO 19 a: Distinguish between the relative assumptions of single and multiple regression.