University of Dhaka

4th Year B.S. (Hons.) In-course Examination 2021

Subject: Statistics, Course No.: Stat H-402
Course Name: Time Series Analysis
Full Marks: 25, Time: 1Hour
Answer all the following questions
I. What do you mean by the term Time Series? Discuss the relationship between S
time series and stochastic process.
2.. 'A strictlystationary time series is also aweakly stationary time series', Justify. 3
3. Show that the time series originated from arandom walk model is not 6
stationary.
4. Suppose Y= Bo + B,t+ X, ,where X; is a zero-mean stationary series with 6
autocovariance function Y, and B's are constants. Show that Y, is not stationary,
but the serics Z, =Y-Y- is stationary.
5
5. The average monthly temperature (in degrces Fahrenheit) over a number of years
recorded in Dubuque, lowa are shown in following figure. Explain the pattem of
the plot.

1972 1974 1976

1066 1070
1084
 4th Year B.S. (Hons.) In-course Examination 2022
Subject: Statistics, Course No.: Stat H-402
Course Name: Time Series Analysis
Time: 1hour Full Marks:25
Answer all the following questions

1. Define general linear process (GLP).Find mean and covariance function for GLP.
5
ind lag-k autocorrelation function (ACF) for AR (1) process.
3. Pind mean and autocorrelation function for second order moving average process:
MA (2).
(P) model"- Justify. 5
4. "Sample PACF is agood indicator of the order of the AR
S. Discuss the problem of overdifferencing,

The End
 Uatverelty of Dhaka
Fourth Year B.N. (Honors) Firnt
Ia-courae Baminstlon, 2021
Nubjeet1 tattstlea
Coursen Net M403
Ting 0 ninutea
FujIMarki 25
1he cuting speeda of four types of toxoh are
being compared in an experiment. Plve
uting materlala of Varying degroe of hardness are to be used for
this purpose. Answer
the folluwing queatlons
6
Mentlon the factor, treatments, experlmental unit, and yleld.
Which exerl1nental dealgn is approprlate here (CRDIRBDILSD?
i. Show the expected layout.
What are the source of experinental efror? How can we control In
Estimate the parameters involved In the flxed effect lincar model for CRD.
4 For four treatments, flnd tho (our) standard l.atin squares. Randomize any one of them.
i. Why is LSD an incomplete throe-way layout?
Comparo CRD, RUD, and LSD in terms of having/reducing eror varietion.
Bonss For comparlng k treatments in the RBD, show that the following two hypoheses are
identical.

i k

i.

where , axd T, be the populatlon mean and the flxed effect of the jth treatment.

TheBnd
 Untverslty of Dhaka
Fourth Year B.S. (Honors) Second In-course Examlnation, 2021
Subjeet: 8tatlstics
Course: Stat H 403
Time: 50 minutes Fall Mrk:25
Find the relative eficiency of CRD relative to RBD. Based on your result, comment (n 7
oneltwo sentences) on the situation when we should conduct experiment in CRD ather
than RBD.
2. "If missing value occurs in RBD,. property of orthogonality will be lasf- prove the 8
theorem.
3. Let we have two factors AandB, both having two levels. Estimate the main effects and 10
show that the interaction effect AB = BA. What is the contribution of cach effect to the

total sum of square?

Bonus Referring to the Question 3, show that the imain effects and interaction efects are three 1
mutually orhogonal contrasts.

The End
 Fourth Yenr B,S. (Hons) First hcourse Bxaminatlon, 2021
Subject: Stntistics
Course No.: Stat H- 404
Course Name: BcnometrlcN
Full Murks: 25, Time: 55 Minutes

Answer the following questions:

1. a) What do you mean by systen of simultancous cquations? Show that (7)
simultaneous relntions provide biascd and inconsistent paramcter
cstimates.
Jb) For the following simultaneous cquation modcl, cxamine the (4)
identification of first structural cquation.
y,=Sy, -2.r, +*, +M

y,= y,-,-3x, + t,

2 Develop a simultaneous equation model for the supply and demand for (5)
Doctors (Child Specialists) in Bangladesh. Specily the endogenous
variables and cxogenous variables in the model.
State with brie> renson whether the following stutenents ure true, falsc,
or uncertain:
a) In the presence of hcteroscedasticity OLS estimators are biased as (2)
well as inefficient.
b) Ifa regressor that has nonconstant variance is (incorrectly) omiltcd (2)
from a model, the (OLS) residuals will be heteroscedastic.
c) lf heteroscedasticity is present, the conventional t and F tcsts ure (2)
invalid.

4. What do you mean by intrinsically nonlincar regrcssion model? Explain (3)

with exanmple. What is the difference betwcen OLS cstimation and non
linear least squares estimation?

THE END.
 Fourth Year B.S. (Hons.)University of Dhaka
Incourse Examination 2021 (Second
Subject: Statistics Incourse)
Course No: Stat I-404
Course Name: Econometrics
Full Marks: 25, Time: 55
Answer the following questions Minutes
1. (a) Discuss any one test
belongs to Farrar-(Glauber test along with the (9)
hypothcsis and decision rule.
(b) Consider the sct of hypothctical data in the
you want to fit the following modcl to the data. following table. Suppose (2+3)

i) Can you estimatë the above three

unknowns? Why or why
not? Explain properly.
ii) If not, what linear functions of these parameters, the
estimable
functions, can you estimate? Show the necessary calculations.
Y -10 -8 -6 -4 -2 2 4 6 10
2 3 4 6 8 9 10
X3 1 3 7 11 13 15 17 19 21

2 Using matrix notation, it can be shown (2+2)

var - cov() = o(X'X)-1
What happens to this var-cov matrix:
a. When there is perfect multicollinearity?
b. When collinearity is high but not perfect?
3 What is logit model? Explain with example. (3)

4 Discuss any one problem of linear probability model. (4)

THHE END.
 In-course I, Department of Statistcs
H-405 (Survival Analysis)
Marks: 25, Time: 50 Minutes
Answer All questions.

Q1. Define hazard function. Based on this definltlon of hazard function, find a 5

relationship between f(t), h(t) and S(t), where notations represent thelr usual
meanings in survival analysis.
Q2. The survival data on myeloma patients are summarized In the following table. 10

Survival time t Number of patlents survivingNumber of patients

(months) beginning at interval dying in interval
40 5
0-5
5-10 35

10-1S 28

15 20 22
5
20-25 18
25-30 13
4
30-35
3540 5
2
40- 45 S
1
45-50 3
2
2
2 50

Determine the estimates off(t)-and S(t)

() h(t).
Using Actuarial method estimate survival function, and hence
find
(i) plot the estimated
estimates in Q2,
Q3. Based on your interpret the value of the
median.
survival time and exponentially distributed random
5
the median function for an
residual life
Q4. Find the mean
variable T.

Page 1 of 1
 In-course 2, Department of Statistics
H-405 (Survival Analysis)
Marks: 25, Time: 50 Minutes
Answer Allquestions.
4 Distinguish between Log-rank test and Wilcoxon test for comparing twO
functions. survival 5
2. Show that if the hazard functions are
proportional, the survivor functions for the two 5
groups of survival data do not cross one another.
3 Find the mean of Weibull distribution.
5
4. Determine the hazard function of Weibull
distribution, and show it graphically when 5
(shape parameter, Scale parameter) = (1, 1) and also when (shape
parameter, Scale
parameter) = (2, 1).
5. Find the median of Extreme value distribution.
5
 H-406 Midternm Exam
Time: 50minutes, Full Marks: 25:
1. What is a stochastic process? Give an example.

2. Explain time-homogeneous and tÉme-nonhomogeneous Markov Chains with

examples.

3. There are three food stores A, B and C in an area. Weekly data on customer
preferences show that, 15% and 10% of customers of Store A go over to B
and C next week, respectively. Similarly, 20% and 25% customers of Store B
go over to A and C, while 10% and 20% customers of StoreC go over to A
and B. Construct the TPM.
/0.1 0.2 0.7
= 1).
4. Let S={1,2, 3} and P=0.2 0.4 0.4 ). Calculate P(X12 =3| X10
\0.1 0.3 0.6/

P given above. Obtain the limiting

5. Refer to the transitional probability matrix
distribution .
The End
 H-406 Midterm Exam
Time: 50 minutes, Full Marks: 25

Classify
1. The transitional probability matrix ofa Markov chain is given below.
the states of the chain.
1 /1/2 0 0 0 0 1/2
22 0 1/3 0 0 2/3 0
3 0.1 0.2 0.3 0.2 0.1 0.1
P=|
4 0 0 0 0 0 1
0 2/3 0 0 1/3 00
5s
6 \1/2 0 0 0 0 1/2/

$1 each day as a
2. You bought a share of stock for $12. Stock price moves
What is the
simple random walk. The probability of an increase is 0.55.
probability that the stock price will reach $15 before reaching $5.
with a rate of
3. Customers arrive at a book store following a Poisson process
the
2.5 per hour. What is the probability that exactly 3 customers will enter
store from 3:30 p.m. to 6:00 p.m.?

4. In a store, there are 20 packets of candy for sale. The sale rate depends on the
number of packets remaining. For one packet, the rate is 0.5/hour. What is
the probability that after 3 hours 5 packets will be remaining?

S. Arrival at a telephone booth forms a Poisson process with a rate of 12/hour.

Length of a telephone call follows exponential distribution with mean 2
minutes. What is the probability that anewcomer has to stay in the system for
more than 2 minutes?

The End
 Fourth Year B.S. (Hons.) Incourse IExamination 2022
Stat H407: Generalized Linear Models
Marks: 25 Time: 50 minutes
Answer all questions. Marks are given in the right margin.

1. (a) Write down the general form of ageneralized linear model (GLM) for the pdfor prnf. 2
(b) Suppose Y~ Bernoulli (). Show that the pmf of Ycan be expressed as the general 3
form of the GLM and hence write down the link function of the GLM with
justification.
2. (a) Wite down the mean and variance properties of a GLM. 2
(b) Suppose Y Poisson (0). Justify the mean and variance properties of a GLM. 3

3. (a) Find the expression for the score statistic in the context of a GLM. 2
(b)Obtain the mean and variance of the score statistic. 6
(c) Wite down the relationship between variances of the score statistic and response
variable.

4. Suppose Y~ Binomial (n, 0). Show that for large sample the score
statistic follows an S
approximately normal distribution with a meán of zero and variance a
 Fourth Yenr B.S. (Ilons.) lncourse II Examination 2021
nt H407: Gencralized Lincar Models
Marks: 25 Time: 50 minutes
Answer all questions.
Binomial(.P), i I,2, ...n und yr s are inlependent. Lct = (Po,...B)"
Nraneter veetor md x; = (X»X{k) is a covariate vcctor of observation i. 2
Writedown the logistic model and hence lind the success probability p,.
(b) Write down the log-likelihood lor B und hence derivc the dcviance.
te) Deline Pearson chi-squared residual and goocdness of fit statistic. Also, write down 3
the deviance residual.
(d) ('onsider the lincur prediclor n= Bo + Bx. IHow do you interpret B,? Justify. 2

2. Individuals who are cxposcd to u toxic subslance are thought to have a probability of
becoming ill which depends on the level of cxposure, x. Data consisting of the response y.
Iif individual i bccame ill, y1 =0 if not, and thc level of exposure, xi, were collected for
n cxposed individuals. logistic regression was fitted, and the maximum ikelihood
estimates were lound to be Bo = -4.25 and B, = 1.75.
(1) Write down the modcl and hence the required equation that needs to be satisfied by 2
the estimated probability function f(x) = P(y = 1|).
(b) Estimate the maximum level of exposure that can be permitted to ensure a
less than or equal to 0.03 that the individual becomes ill. probability 2
(c) Estimale probability of an individual becoming ill when the level of exposure is 0.6.
2
3. Derive the deviance for Poisson regression and hence show the
formulation used in the 7
contingency tables and log-liner models.
 B.S. (Honors) 2nd Midterm
'ourth Year
Course: Stat. H-408
Total Marks: 25 (Comprehensi
Exami n ation, 2021
Time: 1Hourve)
Answer all questions
Definesamplingnit and
sampling frame. Write down the criteria of an
1. ideal (3)
sanpou
What mean simple random sampling
by
(SRS)? When SRS is used? (10)
2.
Supposethat a SRS of size 2 is drawn from a
replacement. Also, suppose that values population of size 4 without
respectively. obtained from sampling units 1,
4are 5, 10, 15, aFd 20, 2,3, and
)Find the population mean,
i) Find the all possible sample with sample values. Also, find the all possible
sample mean values.
ia) Show that sample mean 1S an unbiased
estimator of population mean.
Find the sampling distribution of sample mean. Hence show that it is an
unbiased estimator of population mean.
3. Define stratified sampling. What are the allocation
sampling? Explain Neyman-allocation. procedures used in stratified
(4)
A Distinguish between 1) Type-I error and
Type-II error i) Estimation and (3)
hypothesis testing.
5 A
correlation coeficient of 0.3 is obtained from a sample of 30 pairs. Is this value (5)
consistent at 5 percent level with the assumption that the coefficient of correlation
in the population is 0.50?
 University of Dhaka
Fourth Ycar B.S. (Honors) Midterm Examination, 2021
Subject: Statistics
Course: Stat. H408 (Comprehensive)
Total Marks: 25 Time: I Hour
Answer all questions

. Indicate basic characteristics of different measurement scale used toanalyze statistical (5)
and appropriste graphical
data. ldentify the appropriate scale of measurementHappiness,
technique for the following variables: Economic status, Race, Family size,
Temperature, and Fat consumed.
measures of central tendency. Compare Arithmetic mean (AM), Median (Me); (6)
2. Define data, which measure is appropriate? From theorctical
and Mode (Mo). In skewed
the important conditions or characteristics of a good descriptive
standpoint, what are
statistics?

variable. Write down the properties of probability distribution (5)

3. Define random
function. Find marginal density ofX and Y, where
0<x<2,2 <y <4
Otherwise
0, P[X+Y3]and P[XKN
function. Also,compute
Show that g) &g) are probability
Y<SI2).
different models of upright freezers having 14.5, 165, (6)
An appliance dealer sells three
space, respectively.LetX= the amount of storage space
and 19,4 cubic feet of storage to buy a freezer. Suppose that X has the following
purchased by the next customer
probability mass function: 19.4
14.5 16.5
X 0.2
0.3 0.5
P(X)
Compute B(), E(X) and V(X).capacity Xcubic feet is given by Y= 1SX-7.5. Find the
The price ofa freezer havingpaid by the next customer to buy a freezer.
expected piece and variance
Negative binomial distribution turns into geometric (5)
5. Under what condition the
distribution.
from a polling booth and asking them if they voted
You are surveying people exiting a person voted independent is 20%. What is the
independent. The probability that asked before you can find 5 people who voted
probability that 15 people must be
independent?
 Practical Incourse Examination 2021
Department of Statistics, DU
Total Marks: 10 Course: Stat H-409: Multivariate Analysis
Time: 1 Hour

Answer the following questions:

(You can use R/STATA to solve at most one of the two
questions)
l. Draw a sample of 100 observations from a multivariate normal distribution
1
N(4,2),where
-2 01
'-3, 1,4) is the mean matrix and E= 5 |is the covariance matrix. Then draw
0
scatter plots and boxplots of the variables and discuss the aspects of multivariate data
2. Satellite applications motivated the development of asilver-zinc battery. Table 1
contains failure data collected to characterize the performance of the battery during its
life cycle. Type the data in an Excel/SPSS file and read it in R/STATA as appropriate.

(a) Find the estimated linear regression of on an appropriate ("best") subset of

predictor variables.
(b) Analyse the residuals and comment on your results.
2+3]
Table 1: Battery failure data
Z
Depth of
Churge
rate
Dicharge discharge
(%ofrated
(amps) ps) -ampere hoursy
375 3.13 60,0 40 2.00 101
1.000 3.13 76.8 30 1.99 141
1.000 3.13 60.0 2.00 96
1.000 3.13 60,0 20 1.98 125
L625 3.13 43.2 10 2.01 43
1625 3.13 60.0 20 2.00 16
1.625 3.13 60,0 20 2.02 188
375 5.00 76.8 10 2.01 10
1000 3.00 43.2 10 1.99
1.000 S.00 43.2 30 2.01
5.00 386
1.000 100.0 20 2.00 45
1.625 S.00 76.8 J0 1.99
375 1.25 76,8 10 2.01 76
1000 1.25 43.2 10 1.99 78
1000 1.25 76.8 30 2.00 160
1.000 1.25 60.0 2.00 3
1.625 1.25 43.2 30 1.99 216
1.625 1,25 60.0 20 200 .73
.375 3.13 76.8 30 199 314
375 3.13 60.0 20 2.00 170
 University of Dhaka
Fourth Year B.S. (Honors) Practical In-course Examination, 2021
Subject: Statistlcs
Course: Stat H 409
Title: Statistical Computing VII (Experimental Design Part)
Time: 50 minutes
Full Mark: 10
Suppose you are given the following hypothetical dath obtained from acompletely
five treàtments: A, B, C, D, and E.
randomized design on

A B D E
7 9 6
4 12 6
4 6
9 6 3 12
10 3 7 11 13
11 5 9 12

i. Test whether the treatment means are equal or not stating neceSsary assumptions and the
hypothesis.
ii. Test the hypothesis that the true average yields, under treatment A&D are equal.
iii. Find 95% confidence interval for the difference between the true average yiclds under treatment A
& E.

Bonus In the above experiment, it is supposed that there is no existence oftreatment Band E. Treatment B
is nothing but treatment A, whereas treatment Eis nothing but treatment D, implying we have only three
treatments having a total of 30 yields. Now, test the hypothesis that the true average yields under treatment
A &D are equal.

The End
 Stat H-41o (Biostatistlcs)
Marks: 10, Time: 1 hour
Answer AIl
Wite Rcodes where required and only relevant output.
4
1. Conslder the following survival times (in days) obtalned from an ovarlan cancer
study.
59, 115, 156, 421+, 431, 448+, 464,; 475, 477+, 563, 638, 744+
errors.
Find the K-M estimates along with thelr standard
2. Let us simulate a random sample of size 1000 froma weibull dlstribution with shape 3
parameter 2 and scale parameter 1.5 using following Rcodes.
set.seed(your roll number)
rweibull(1000, 2, 1.5)
Now based on this sample find MLEs of the parameters.
3
3. The lifetimes (in months) of 10 pieces of equipment are given below
Item No. 1 2 3 4 5 6 7 8 9 10

Ti 3 50 30 27| 15 25 5
80 70 71| 5540 3030 32 27| 22
distribution. Under the set-up of
Assume that the lifetimes follow an exponential
the parameter of exponential
Type I censoring find a 95% confidence interval of
distribution using any method discussed in the class.
 Untversity of Dhaka
Fourth Year B.S. (Honors) In-course Examination 2021
Subject: Statlstics
Course: Stat H 410 (Practica: Time Serles Analysis)
Time: 1 hour Full Marks: 10
Use the datasct "unemp.txt" located in "p" drive o answer the following questions:

1. Convert the data in time series using command., ts(data, freq=12, start=c(1961,1). Plot the data and comment on
stationarity.
2. Use Augmented Dickey-Fuller (ADF) test and verify your answer in question (1).
3. If your answer in ()is "non-stationary", then use appropriate transformation to make the series stationary. Use
ADF test to show that the transformed data is "stationary".
4. Draw ACF and PACF, and suggest an appropriate model. Write down the model in ARIMA(Pd,g) notation.
s. Perform model diagnostic and comment whether your suggested model is adequate.

Useful Code

Data=scan("path"); model-arima(data,order=c(p,d,q)); plot(data); tsdiag(model), adf.test(data), acf<data)

 Fourth Year B.S. (Ilons) IncourseExamination, 2021
Subject: Statistics
Course No.: Stat H- 411
Course Name: Satistical Computing IX: Feonometrics (Using R/SAS/SPSSISTATA)
Pull Marks: 10, Time: 55 Minutes

Answer the following qucstions:

W'rite down stata code in each case.

have given a data named "workstatus". The variables are as (4)

1. You
follows:
workstatus takes
Workstatus means. if the person is working, then
person is not working, then workstatus takes value 0.
value 1. if the
gpa are income of the respondents and gpa of the respondents
Income.
Rank is a categorical variable and it has 4 categories.
respectively.
1means highest ranking under grade educational institute, from
Rank
respondents were graduated, rank 2 means comparing with
where the means comparing
quality educational institute, rank 3
1. little bit less institute and 4 means comparing
with l and 2, less quality educational
educational institute. Estimate the logit
with 1. 2 and 3, less quality
estimated logit model and interpret the results.
model. Write down the

income data". Check the data for: (3+3)

named,
2. You have given a data present then use remedial measure to remove
i) autocorrelation and if
autocorrelation.
residual versus fitted plot and make a comment.
ii) Draw
iii)Verifyyour result obtained from ii).

THE END.
 Fourth Year B.S. (Hons.) Incourse (Practical) Examination 2021
Stat H411:
Marks: 10 Generallzed Linear Models
Time: 50 minutes
Answer all questions.
Considerthe data set medpar(Hilbe, 2011) and our main
goal is to investigate the effect
of died (l: died; 0: alive) and white (1: white; (0: non-white) on thc los
spent in the hospital [number of
(length-of-stay
days]).
1. Write down the model fitted with clear
notation, stating all assumptions,
and identify clearly response variable, and
2. Fit a suitable GLM and hence write down the covariates.
link function and R-code.
3. Predict the average length of stay for the person who is non-white and
alive.
4. Conduct an overall significance test (change of deviance test) for the model
mentioning null and alternative hypotheses at 5% level.
5. Compute incidence rate ratio (IRR)for all covariates and hence interpret
the results.
 First In-course for Fourth Year BS
(Honors) Examination, 2021
Subject: Statistics
Course: H-401: Multivariate Analysis
Time: 1 hour Full Marks: 12.5

Answer the following questions

1. Define multivariate analysis with real life example. How do you organize your
multivariate data?
[1+1]

pefine multinomial distribution with examples Derive the moment generating function of
it and find its first two moments.
(1+3]
3. If X1, Xn be a random sample from a population with meán and covariance
matrix E, then show that û = X and =(X -X)(X,- X) = Sare the
maximum likelihood estimators of and Zrespectively, where, =
S 1X n-1

)(K, ). [3.5]

4. IfX X, ..X be a random sample from N(4,2). Then find the sampling distribution
of Ay = j=(X; - ~)X, -).
[3]
 Fourth Year BS Hons Second Incourse Examination 2021
Subject: Statistics
Course: Stat H-401: Multivariate Analysis
Time: I hour
Full Marks: 12.5
Answer the following questions

Discuss general likelihood ratio method for testing multivariale vector mcans briefly. Also find the
confidence regions and simultancous comparisons of component mcans. [4)

2. Discuss how you can compute the maximum likelihood cstimates from incomplete data. (3)

VS. Discuss bricfly on testing for equality of treatments in a repcated measures design. (3]

A. Stalc the sieps of simultaneous confidence intervals for treatment effects in a one-way MANOVA. (2.5)

(2+3]
Table 1: Battery failure data
Z
Depth of End of
Charge
zate
Discharge
rale
disch charge
raled Tenperature avolege Gyolo10
(amps) (amps) ampere-hours) falltc
375 3.13 60.0 40 200
1.000 101
3.13 76.8 30 1.99 141
1.000 3.13 60.0 20 2.00 96
1000 3.13 60.0 20 1.98 125
l625 3,13 43.2 10 2.01 43
1.625 3.13 60.0 20 2.00 16
1.625 3.13 60.0 20 2,02 188
375 S.00 76.8 10 2.01 10
1.000 5.00 43.2 10 1.99 3
1.000 5.00 43.2 30 2.01 386
1.000 5.00 100.0 20 2.00 45
1625 5.00 76.8 1.99 2
.375 I.25 76,8 10 2.01 76
1000 I.25 43.2 10 1.99 78
1000 1.25 76.8 30 2.00 160
1.000 1.25 60,0 0 2.00 3
1625 1.25 43.2 30 1.99 216
1.625 1.25 60.0 20 2.00 73
375 3.13 76.8 30 1.99 314
375 3.13 60.0 20 2.00 170