Isi JRF Stat 07
Isi JRF Stat 07
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.
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
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:
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
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
Xk2 1
E( 2 ) = , ∀ 1 ≤ k ≤ j.
Vj j
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
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 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
6
Sample Questions : RSII
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→∞
(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 ?
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