Test Code: RSI/RSII (Short Answer Type) 2007

Junior Research Fellowship for Research Course in Statistics

The candidates for research course in Statistics will have to take two short-answer
type tests – RSI and RSII. Each test is of two-hour duration. Test RSI will have
about 10 questions of equal value, set from different topics in Mathematics at B.
Sc (Pass) level and Statistics at B.Sc. (Honours) level. Test RSII will have roughly
8 questions on topics in Statistics at Master’s level.

Syllabus for RSI and RSII

Mathematics
Functions and relations. Matrices - determinants, eigenvalues and eigenvectors,
solution of linear equations, and quadratic forms.
Calculus and Analysis - sequences, series and their convergence and divergence;
limits, continuity of functions of one or more variables, differentiation, applications,
maxima and minima. Integration, definite integrals, areas using integrals, ordinary
linear differential equations.
Statistics
(a)Probability: Basic concepts, elementary set theory and sample space, conditional
probability and Bayes theorem. Standard univariate and multivariate distribu-
tions. Transformations of variables. Moment generating functions, characteristic
functions, weak and strong laws of large numbers, convergence in distribution and
central limit theorem. Markov chains. Birth and death processes.
(b) Inference: Sufficiency, minimum variance unbiased estimation, Bayes estimates,
maximum likelihood and other common methods of estimation,. Optimum tests for
simple and composite hypotheses. Elements of sequential and non-parametric tests.
Analysis of discrete data - contingency chi-square.
(c) Multivariate Analysis: Standard sampling distributions. Order statistics with
applications. Regression, partial and multiple correlations. Basic properties of
multivariate normal distribution, Wishart distribution, Hotelling’s T-square and
related tests.
(d) Linear Models and Design of Experiments: Inference in linear models. Standard
orthogonal and non-orthogonal designs. Inter and intra-block analysis of general
block designs. Factorial experiments. Response surface designs. Variance compo-
nents estimation in one and two-way ANOVA.
(e) Sample Surveys: Simple random sampling, Systematic sampling, PPS sampling,
Stratified sampling. Ratio and regression methods of estimation. Non-sampling
errors, Non-response.

1
 Sample Questions : RSI

1. Let A and B be two given subsets of Ω. For each subset C of Ω define


1 if ω ∈ C
Ic (ω) =
0 if ω ∈ / C.

Does there exists a subset C of Ω such that | IA (ω) − IB (ω) |= Ic (ω) for each
ω ∈ Ω ? Justify your answer.
2. (a) Evaluate the determinant
 
 1 + x1 y1 1 + x1 y2 ... 1 + x1 yn 
 
 1 + x2 y1 1 + x2 y2 ... 1 + x2 yn 
.
 
 .. .. ..

 . . . 

 1 + xn y1 1 + xn y2 . . . 1 + xn yn 

(b) Find the inverse of the matrix A + αα0 , where A is a diagonal matrix
diag(a1 , a2 · · · , ak ) and α0 = ( α11 , · · · , α1k ), αi > 0 for i = 1, · · · , k.
3. Let An×m and Bm×n be two real matrices. Show that the nonzero eigenvalues
of AB and BA are the same.
4. Let f , g and h be defined on [0, 1] as follows:

f (x) = g(x) = h(x) = 0 whenever x is irrational;

f (x) = 1 and g(x) = x whenever x is rational;
1
h(x) = if x is the rational number m/n (in lowest terms);
n
h(0) = 1.

Prove that
(a) f is not continuous anywhere in [0, 1],
(b) g is continuous only at x = 0, and
(c) h is continuous only at the irrational points in [0, 1].
5. Consider a function f which satisfies the relations :
1
2f 0 (x) = f ( ) if x > 0, f (1) = 2.
x
(a) Let y = f (x). Find suitable values for constants a and b such that y
satisfies a differential equation of the form x2 y 00 + axy 0 + by = 0.

2
 (b) Determine a solution of the above differential equation of the form f (x) =
cxn .
6. Let f : < → < be differentiable at α. Further, f (α) = 0 and let g(x) = |f (x)|.
Show that g is differentiable at α if and only if f 0 (α) = 0.
Z Z
7. Find the maximum of dxdy as a function of m for 0 < m < 1 where
4

x2 y2
4 = {(x, y) : + ≤ 1}
m 1−m
R4
8. (a) Find −2
[x]dx, where [x] is the largest integer less than or equal to x.
(b) If f (x), x ≥ 0 is a differentiable function such that f 0 (x) → ∞ as x → ∞,
then show that f (x) → ∞ as x → ∞.
9. (a) Find the limit
n
X n
lim
n→∞ n2 + k 2
k=0

(b) Let f (t) be a continuous function on (−∞, ∞) and let

Z x
F (x) = (x − t)f (t)dt. Find F 0 (x).
0

10. (a) Find max(xyz) subject to x2 + 2y 2 + 9z 2 = 6.

Pn Pn ak
(b) Let f (x) = k=0 ak xk , where ak ’s satisfy k=0 k+1 = 0.
Show that there exists a root of f (x) = 0 in the interval (0, 1).
11. One of the sequences of letters AAAA, BBBB, CCCC is transmitted over a
communication channel with respective probabilities p1 , p2 , p3 , where (p1 +
p2 + p3 = 1). The probability that each transmitted letter will be correctly
understood is α and the probabilities that the letter will be confused with two
other letters are 12 (1 − α) and 12 (1 − α). It is assumed that the letters are
distorted independently. Find the probability that AAAA was transmitted if
ABCA was received.
12. In a sequence of independent tosses of a fair coin, a person is to receive Rs.2
for a head and Re.1 for a tail.
(a) For a positive integer n, find the probability that the person receives
exactly Rs. n at some stage.
(b) What is the limit of this probability as n → ∞ ?

3
13. Let X1 , X2 , · · · , Xn be i.i.d. random variables with P (Xi = −1) = P (Xi =
1) = 12 . Let Sn = X1 +· · ·+Xn for n ≥ 1. Let N = min{n ≥ 1, such that Sn =
0}.
(a) Show that P (N = 2k + 1) = 0 for every integer k ≥ 0.
(b) Find P (N = 2k, Sk = k) for every integer k ≥ 1.
14. (a) Show that if X1 and X2 are identically distributed (not necessarily inde-
pendent) random variables then, for t > 0,
t
P [|X1 − X2 | > t] ≤ 2P [|X1 | > ].
2

(b) If X1 , X2 , . . . are independent and identically distributed random vari-

1
ables and Vj = (X12 + X22 + . . . + Xj2 ) 2 , then show that

Xk2 1
E( 2 ) = , ∀ 1 ≤ k ≤ j.
Vj j

15. Let X1 , · · · , Xn , · · · , be√a sequence of independent random variables with

E(Xr ) = 0, Var(Xr ) = r for r = 1, · · · , n, · · ·. Prove that n1 (X1 + · · · + Xn )
converges almost surely and find the limit.
16. Suppose die A has 4 red faces and 2 green faces while die B has 2 red faces
and 4 green faces. Assume that both the dice are unbiased. An experiment is
started with the toss of an unbiased coin. If the toss results in a Head, then
die A is rolled repeatedly while if the toss of the coin results in a Tail, then
die B is rolled repeatedly. For k = 1, 2, 3, · · · , define

1 if the kth roll of the die results in a red face
Xk =
0 otherwise.

(a) Find the probability mass function of Xk .

(b) Calculate ρ(X1 , X7 ).
17. A model that is often used for the waiting time X to failure of an item is
given by the probability mass function

pX (k|θ) = θk−1 (1 − θ), k = 1, 2, . . . , 0 < θ < 1.

Suppose that we only record the time to failure if the failure occurs on or before
r, and otherwise just note that the item has lived at least (r+1) periods. Let
Y denote this censored waiting time.
(a) Write down the probability mass function of Y .
(b) If Y1 , Y2 , . . . , Yn is a random sample from the censored waiting time, write
down the likelihood function and find the MLE of θ.

4
18. Let X be exponentially distributed with mean 1 and let U be a U [0, 1] random
variable, independent of X. Define

I = 1 if U ≤ e−X and I = 0 otherwise .

Show that the conditional distribution of X given {I = 1} is exponential with

mean 0.5.
19. Let X be an exponential r.v. with mean 12 . Let Y be the largest integer less
than or equal to X. Find the probability distribution of Y .
20. Let X1 , X2 , X3 be independent standard normal variables. Find the distribu-
tion of
X1 + X2 + X3
U= .
X2 − 2X3 + X1
21. Let Y have a chi-squared distribution with k(≥ 3) degrees of freedom. Express
E(1/Y ) in terms of k.
22. Let X follow an Np (θ, Σ) distribution, where Σ is a known positive definite
P−1
matrix and θ is either θ1 or θ2 , θi ’s being known. Show that (θ2 − θ1 )0 X
is a sufficient statistic.
23. Suppose X1 , X2 , . . . , Xn is a random sample from a population with proba-
bility density
x x2
p(x|θ) = 2 exp(− 2 ), x > 0, θ > 0.
θ 2θ
Find the Fisher information number I(θ) and give a lower bound to the vari-
ance of an unbiased estimator of θ.
24. Show that X1 + 2X2 is not sufficient for µ where X1 and X2 are two random
observations from N (µ, 1) distribution.
25. The number of accidents X per year in a manufacturing plant may be mod-
elled as a Poisson random variable with mean λ. Assume that accidents in
successive years are independent random variables and suppose that you have
n observations.
(a) How will you find the minimum variance unbiased estimator for the prob-
ability that in a year the plant has at most one accident ?
(b) Suppose that you wish to estimate λ. Suggest two unbiased estimators
of λ and hence find the UMVUE of λ.
26. Let X1 , X2 , · · · , Xn be a random sample of size n from a negative exponential
distribution with mean θ. The experiment is terminated after the first r(r ≤
n) smallest observations have been noted.
Write down the likelihood for θ based on these censored observations. Find
the mle of θ. Obtain the UMP test for testing H0 : θ = θ0 vs. H1 : θ > θ0 at
level α.

5
27. Consider the following fixed effects linear model:

Y1 = θ1 + θ2 + θ4 + 1
Y2 = θ1 + θ3 + 2θ4 + 2
Y3 = θ1 + θ2 + θ4 + 3
Y4 = −θ2 + θ3 + θ4 + 4
 = (1 , 2 , 3 , 4 ) ∼ N (0, σ 2 I4 )

(a) Find the full set of error functions.

(b) Hence or otherwise, find an unbiased estimator of σ 2 with maximum
possible degrees of freedom.
28. An experimenter wanted to use a Latin Square Design (LSD) but instead,
used the following row-column design

A B C
B A C
C A B

Are all treatment contrasts estimable ? Give reasons for your answer.
29. An unbiased coin is to be used to select a Probability Proportional to Size
With Replacement (PPSWR) sample in 2 draws from the following population
of 3 units, where X is the size measure.

i 1 2 3
Xi 2 4 1

Consider the following procedure.

a) Toss the coin thrice independently.
b) If the outcome is {HHH} or {HHT }, select the first unit.
c) If the outcome is {HT H}, {HT T }, {T HH} or {T HT }, then select the
second unit.
d) If the outcome is {T T H}, then select the third unit.
e) If the outcome is {T T T }, do not select any of the units, go back to (a)
above.
f) Continue till a unit is selected.
Show that the above procedure is a PPS method. If W = number of tosses
required to select a unit, find E(W ).
30. Show that under PPSWOR of size 2 or 3, inclusion probability of ith unit
exceeds that of jth unit, whenever ith unit has a larger size measure than
unit jth unit.

6
 Sample Questions : RSII

1. Express the following as the probability of an event and evaluate without

integrating by parts: Z ∞
1 x2
Φ(x) √ e− 2σ2 dx,
−∞ 2πσ
where Φ(·) is the standard normal cumulative distribution function.
2. Suppose that X1 , X2 , X3 , · · · , are i.i.d. random variables with EX1 = 0,
EX12 = 1, EX14 < ∞. Show that
n
X
n−1/2 [ (Xi − X n )2 − n]
1

Pn in law to a normal distribution with zero mean, as n → ∞. Here

converges
X n = 1 Xi /n.
3. Let X1 , X2 , · · · , Xn be i.i.d. observations from a continuous distribution and
R1 , · · · , Rn be the ranks of the observations. Find Cov(Rn−1 , Rn ).
4. Let {Xn } be a sequence of independent random variables. Show that for each
real number α, P (Xn → α) = 0 or 1 by explicitly applying the Borel-Cantelli
lemmas.
5. Let (Xi , Yi ), i = 1, 2, · · · , n be a random sample on (X, Y ) which is a bivariate
vector of the continuous type. Let T = Max.(X, Y ) and Ri be the rank of
Pi namong T1 , T2 , · · · , Tn . Find the mean and variance of the statistic S =
T
i=1 (n+1−Ri )δi assuming that (X, Y ) and (Y, X) are identically distributed,
where δ = I{X<Y } is the indicator function.
6. A company desires to operate S identical machines. These machines are sub-
ject to failure according to a given probability law. To replace these failed
machines the company orders new machines at the beginning of each week to
make up the total S. It takes one week for each new order to be delivered.
Let Xn be the number of machines in working order at the beginning of the
nth week, and let Yn denote the number of machines that fail during the nth
week.
(a) Establish the recursive formula Xn+1 = S −Yn , and show that Xn , n ≥ 1
constitutes a Markov Chain.
(b) Suppose that the failure rate is uniform i.e.,
1
P [Yn = j | Xn = i] = , j = 0, 1, · · · , i.
i+1
Find the transition matrix of the Chain, its stationary distribution, and
expected number of machines in operation in the steady state.

7
 7. Let Nn be the number of heads observed in the first n tosses of a fair coin,
with N0 = 0. Find
lim P [Nn is a multiple of 3]
n→∞

by formulating the problem in relation to a Markov Chain with state space

{0, 1, 2} consisting of the remainders obtained from non-negative integers after
division by 3.
8. (a) In a forest there are 50 tigers and an unknown number L of lions. Assume
that the 50 + L animals randomly move in the forest. A naturalist sights
5 different lions and 15 different tigers in the course of a trip in the forest.
Estimate, stating assumptions and with theoretical support, the number
of lions in the forest.
(b) A bank decided to examine a sample of vouchers before conducting a
thorough audit. From a very large number of accumulated vouchers,
samples were drawn at random. The first defective voucher was obtained
in the 23rd draw, and the second defective voucher in the 57th draw.
Estimate, giving reasons, the proportion of vouchers that are defective.
9. Let X and Y be two random variables such that

P (X > x, Y > y) = exp[−λ1 x − λ2 y − λ12 max(x, y)].

(a) Find the marginal distribution of X.

(b) Let λ12 = 0.5. Find the method-of-moments estimators of λ1 , λ2 .
(c) Under independence of X and Y in the above family of distributions and,
based on a sample of n pairs of observations (Xi , Yi ) from the resulting
distribution, find an estimator of λ1 /λ2 .
(Hint: E(χ2ν ) = 1/(ν − 2), ν > 2.)
10. Let Y(1) < Y(2) < · · · < Y(n) be the ordered random variables of a sample of
size n from the rectangular (0, θ) distribution with θ unknown, 0 < θ < ∞. By
a careless mistake the observations Y(k+1) , · · · , Y(n) were recorded incorrectly
and so they were discarded subsequently (Here 1 ≤ k < n).

(a) Show that the conditional distribution of Y(1) , · · · , Y(k−1) given Y(k) is
independent of θ.
(b) Hence, or otherwise, obtain the maximum likelihood estimator of θ and
show that it is a function of Y(k) .
(c) If nk → p as n → ∞, for some 0 < p < 1, what can you say about the
asymptotic distribution of the maximum likelihood estimator of θ ?

11. Let X ∼ N6 (µ, Σ), where µ = (µ1 , · · · , µ6 )T , and Σ is unknown. Obtain an

exact test for testing H0 : µl = lη, l = 1, 2, · · · , 6, η being unknown, against
H1 : not H0 . Obtain the cut-off point of this test.

8
12. The value of Y is estimated from X = x0 and the linear regression of Y on X.
Let this estimated value of Y be y0 . Then the value of X corresponding to
Y = y0 is estimated from the linear regression of X on Y . Let this estimated
value of X be x∗0 . Compare x0 and x∗0 . Interpret your answer.
13. Let X be a random variable having a density θ1 e−x/θ , x, θ > 0. Consider
H0 :Pθ = 1 vs. H1 : θ = 2. Let ω1 and ω2 be two critical regions given by
n
ω1 : 1 Xi ≥ C1 and ω2 : ( number of Xi ’s ≥ 2) ≥ C2 .
(a) Determine approximately the values of C1 and C2 for large n so that
both tests are of size α.
(b) Show that the powers of both tests tend to 1 as n → ∞.
(c) Which test would require more sample size to achieve the same power ?
Justify your answer.
14. Let X1 , X2 , . . . , Xn be i.i.d observations with a common exponential distri-
bution with mean µ. Show that there is no uniformly most powerful test for
testing H0 : µ = 1 against HA : µ 6= 1 at a given level 0 < α < 1 but there
exists a uniformly most powerful unbiased test and derive that test.
15. Let X = (X1 , X2 , X3 , X4 ) have a multivariate normal distribution with un-
known mean vector µ = (µ1 , µ2 , µ3 , µ4 ) and unknown variance covariance
matrix Σ which is nonsingular. Let X1 , · · · , Xn be a random sample of size n
from the population. Develop a test for testing the hypothesis H0 : µ1 +2µ2 =
µ2 + 2µ3 = µ3 + 2µ4 . State the distribution of the test statistic.
16. Consider the following linear model under the usual Gauss-Markov setup:

E(y1 ) = β0 + β1 + β2
E(y2 ) = β0 + β1 + β3
E(y3 ) = β0 + β2 + β3
E(y4 ) = β0 + β1 + β2
3
X
(a) Determine the condition of estimability of the parametric function li βi ,
i=0
where l0 , l1 , l2 , l3 are known constants.
(b) Obtain the best linear unbiased estimator of (2β1 − β2 − β3 ) and also
obtain its variance.
17. Consider the following 24 factorial design with factors A, B, C and D in the
usual order.
Block 1 : (0,0,0,0), (0,1,0,1), (1,0,1,0), (1,1,1,1), (1,1,0,1), (0,0,1,0).
Block 2 : (0,0,1,1), (0,1,1,0), (1,0,0,1), (1,1,0,0), (1,1,1,0), (0,0,0,1).
Block 3 : (0,1,0,0), (0,1,1,1), (1,0,0,0), (1,0,1,1).
Examine whether the main effect of A is estimable.

9
18. Let (5, 0.023), (3, 0.189), (5, 0.023) and (2, 0.272) denote the (i, pi )-values for
the labels i selected in 4 draws with replacement from a population of size N
with selection probabilities pi on each draw. Give (a) an unbiased estimate
of N and (b) indicate, without actual evaluation, how a suitable confidence
interval for N could be set up.

19. A sample s(1) of n units is selected from a population of N units using SR-
SWOR and the values of a variable y are ascertained for those n1 units among
the n units of s(1) who responded. Later, a further sub-sample s(2) of m units
is selected using SRSWOR out of the (n − n1 ) units of s(1) which did not re-
spond. Assuming that the y-values of all the m units of s(2) could be obtained,
find the following :
(a) an unbiased estimator of the population mean Y on the basis of available
y-values.
(b) an expression for the variance of the proposed estimator.
20. Let the size of a population be N = 3 and the sample size n = 2. Let
s1 = (1, 2), s2 = (1, 3), and s3 = (2, 3) denote the three possible samples.
Under simple random sampling, you have p(si ) = 1/3, i = 1, 2, 3. Define the
estimator t by
y +y

 t1 = 1 2 2 if s1 occurs
t= t2 = y21 + 2y33 if s2 occurs
t3 = y22 + y33 if s3 occurs.


Show that t is unbiased for Ȳ and that there exist values (Y1 , Y2 , Y3 ) for which
V (t) < V (ȳ), where ȳ denotes the conventional sample mean. What does this
example imply?

10