Mains PYQs
UPSC IFoS Mains Statistics paper 1 2019 questions
- An unbiased six-sided die is thrown twice. Let $X$ denote larger of the scores obtained. Then show that the probability mass function (p.m.f.) is given by
$$p_X(x) = \begin{cases} \frac{2x-1}{36}, & x = 1, 2, \ldots, 6 \\ 0, & \text{elsewhere} \end{cases}$$
- If $X$ follows the binomial $(n, p)$, then for any $a > 0$, show that
$$P\left[\left|\frac{X}{n} - p\right| \geq a\right] \leq \frac{1}{4na^2}$$
- Let $X_k$, $k = 1, 2, 3, \cdots$ be i.i.d. exponential random variables with mean $\lambda$. Find the limiting distribution of first-order statistic $X_{(1)}$ of sample of size $n$.
- Let $X$ belong to an exponential family of distributions of the form
$$f_X(x, \theta) = e^{x\log\theta + \log g(\theta) + w(x)}$$
Obtain maximum likelihood estimation equation based on $n$ observations. Solve the same for Poisson distribution with variance $\theta$.
- Find a uniformly minimum variance unbiased estimator for the parameter $\lambda$ based on a random sample of size $n$ drawn from the distribution
$$f(x, \theta) = \begin{cases} \sqrt{\frac{\lambda}{\pi}}\, x^{-\frac{1}{2}} e^{-x\lambda}, & x > 0;\ \lambda > 0 \\ 0, & \text{otherwise} \end{cases}$$
- Let $X_1, X_2, \cdots, X_n$ be random sample on $X$ following $U(0, \theta)$. Show that
$$\phi_1(t) = \begin{cases} 1, & \text{if } t > \theta_0 \\ \infty, & \text{if } 0 < t < \theta_0 \end{cases}$$
is most powerful size $\alpha$ test for testing $H_0 : \theta = \theta_0$ against $\theta = \theta_1 > \theta_0$, where $T = X_{(n)}$, $n$th order statistic. Further, show that
$$\phi_2(t) = \begin{cases} 1, & \text{if } t > \theta_0 \\ v(t), & \text{if } 0 < t < \theta_0 \end{cases}$$
gives class of size $\alpha$ tests which have same power as $\phi_1(t)$ subject to the conditions that
$$\int_0^{\theta_0} v(t)\frac{nt^{n-1}}{\theta_0^n}\,dt = \infty$$
and $0 \leq v(t) < 1$.
- Prove that for any $n (\geq 2)$ events $A_j$, $j = 1, 2, \cdots, n$
$$\sum_{j=1}^{n} P(A_j) \geq P\left(\bigcup_{j=1}^{n} A_j\right) \geq \sum_{j=1}^{n} P(A_j) - \sum_{1 \leq j \leq k \leq n} P(A_j \cap A_k)$$
- Let $\{X_n\}$ be a sequence of random variables such that
$$P[X_n = i] = \begin{cases} \frac{1}{n}, & \text{if } i = 1 \\ 1 - \frac{1}{n}, & \text{if } i = 0 \\ 0, & \text{otherwise} \end{cases}$$
Show that $X_n \to 0$ in probability, but not almost surely.
- Let distribution of $X$ belong to one-parameter exponential family of distributions with probability function
$$f_X(x, \theta) = e^{u(\theta)T(x) + v(\theta) + w(x)}$$
Then show that $T(X)$ is unbiased for $\psi(\theta) = -\frac{v'(\theta)}{u'(\theta)}$. Further, show that it is minimum variance bound estimator.
- Let $X$ follow a geometric distribution with probability function
$$P[X = k] = \left(\frac{a-1}{a}\right)^k\left(\frac{1}{a}\right),\ k = 0, 1, \cdots$$
and $a$ is a positive integer $\geq 2$.
(i) Compute the value of $P[|X - E(X)| \geq a(a-1)]$. [7]
(ii) What is the upper bound for the above probability? [8]
- Let $X_1, X_2, \cdots, X_n$ be a random sample of $X$ following $U(0, \theta)$. Show that joint likelihood has a monotone likelihood ratio in $n$th order statistic of the sample.
- For a certain industrial firm, it is found that if a person joins the firm, the probability that he is still there at times $x$ later, where $x$ is measured in years, is
$$\frac{1}{3}e^{-x/2} + \frac{2}{3}e^{-x/4}$$
Find the probability that a person spends between 2 and 4 years with the firm, and also find the mean and variance of length of time he works there.
- A traveller bus operator has 48-seater passenger buses and 38-seater passenger buses. With $X$ and $Y$ denoting number of miles travelled per day for the 48-seater passenger bus and the 38-seater passenger bus respectively, the bus operator is interested in testing the equality of the distributions $F(x)$ and $G(y)$ of $X$ and $Y$ respectively. That is $F(z) = G(z)$ for all $z$ against $F(z) \neq G(z)$ for at least one value of $z$.
The company observed the following data on random sample $n_1 = 10$ of 48-seater and $n_2 = 11$ of 38-seater buses :
$X$ : 104, 253, 300, 308, 315, 323, 331, 396, 414, 452
$Y$ : 184, 196, 197, 248, 260, 279, 355, 386, 393, 432, 450
Using the normal approximation to run test, carry out run test for the problem by considering $\alpha = 0.05$. The table value $z_{\alpha/2} = 1.96$.
- Describe likelihood equivalence of data points. Let $X_1, X_2, \cdots, X_n$ be a random sample on $X$ following $N(\mu, \sigma^2)$. Further, assume that $\sigma^2 = |\mu|$. Obtain minimal sufficient statistic for $\mu$.
- Explain the need for sampling with varying probabilities. Describe a method of selecting a sample with probability proportional to size.
- Why is it said that a split-plot design confounds main effects? Give an outline of analysis of a split-plot design.
- Suppose it is required to estimate the average value of output of a group of 5000 factories in a region so that the sample estimate lies within 10% of the true value with a confidence coefficient of 95%. Determine the minimum sample size required. The population coefficient of variation is known to be 60%.
- A linear model is specified by $E(Y) = X\beta$, where $\beta = (\theta_1, \theta_2, \theta_3)^T$, $Y = (Y_1, Y_2, Y_3, Y_4)^T$ and
$$X = \begin{pmatrix} 1 & 0 & -1 \\ 1 & 1 & -1 \\ 1 & 0 & -1 \\ 1 & -1 & -1 \end{pmatrix}$$
(i) Check whether all parametric functions are estimable.
(ii) Find an estimator for $\theta_2$.
- Given the correlation matrix
$$P = \begin{bmatrix} 1 & 0.5 & -0.3 \\ 0.5 & 1 & -0.47 \\ -0.3 & -0.47 & 1 \end{bmatrix}$$
compute multiple correlation coefficient $R_{1(23)}$ and partial correlation coefficient $r_{13.2}$.
- Under the assumption that the values of the units in population are steadily increasing by a constant amount, show that
$$\mathrm{Var(st)} < \mathrm{Var(sy)} < \mathrm{Var(SRS)}$$
- A system of partial confounding in a $2^3$-experiment in block of $2^2$ is
| Replicate-I Control Block | Replicate-II Control Block | Replicate-III Control Block | Replicate-IV Control Block |
| --- | --- | --- | --- |
| (1) | (1) | (1) | (1) |
| ab | ac | bc | ab |
| c | b | a | ac |
| abc | abc | abc | bc |
Identify confounded effects in different replicates. Considering $r$-replications of this basic model, write down the ANOVA table.
- What purpose does discriminant analysis serve? Explain how Fisher discriminant function is useful in this analysis.
- Find the mean vector and dispersion matrix of the normal distribution given below :
$$f(x, y) \propto e^{-\frac{1}{2}\{2x^2 + y^2 + 2xy - 22x - 14y + 65\}},\ -\infty < x, y < \infty$$
- Let $X \sim N_p(\mu, \Sigma)$ and $X = \begin{pmatrix} X^{(1)} \\ X^{(2)} \end{pmatrix}$, where $X^{(1)}$ is a $q \times 1$ and $X^{(2)}$ is a $(p-q) \times 1$ vector of random variables. Obtain the marginal density of $X^{(2)}$ and the conditional expectation of $X^{(1)}$, given $X^{(2)} = x^{(2)}$.
- Prove that the number of treatments common between any two blocks in a symmetrical BIBD is $\lambda$.
- (i) If $X \sim N_p(\mu, \Sigma)$, then show that $(X - \mu)^T\Sigma^{-1}(X - \mu)$ is distributed as chi-square with parameter $p$. [9]
(ii) State and explain any two uses of $T^2$-statistic. [6]
- Define Hotelling's $T^2$-statistic. Show that it is invariant under the transformation of scale.
- A sample of 30 students is to be drawn from a population consisting of 300 students belonging to two colleges $A$ and $B$. The mean and SD (standard deviation) of their marks are given below :
| | Total number of students ($N_i$) | Mean ($\bar{y}_{N_i}$) | SD ($\sigma_i$) |
| --- | --- | --- | --- |
| College A | 200 | 30 | 10 |
| College B | 100 | 60 | 40 |
Determine the sample sizes in two colleges by using proportional allocation technique. Hence, obtain variance of estimate of the population mean.