Mastering Discrete Probability Distributions & Sampling Theory: Symmetry, Transformations, and Variance Shortcuts (Deep Dive + 10 Practice Sets + 3 Killer Challenges)
[Discrete Probability & Sampling Theory] Beyond Basic Tables: Symmetry, Variable Transformations, and Variance Shortcuts
Statistics is not about tedious arithmetic. By mastering distributional symmetry, composite variable transformations, and the variance scaling of the sample mean ($\overline{X}$), even the most challenging high-tier questions can be solved in three clean steps.
1. Introduction: 3 Core Mechanics for High-Scoring Probability & Statistics
Competitive exams test your structural intuition, not your ability to fill out tables mechanically. Here are the three structural perspectives every top-tier student must possess:
When a probability mass function (PMF) exhibits bilateral symmetry ($P(c-x) = P(c+x)$) or equal spacing, the expected value $E(X)$ equals the axis of symmetry immediately. Translating the variable via $X - \mu$ reduces the computational overhead of $E(X^2)$ and $V(X)$ by more than half.
When dealing with absolute values ($|X|$), truncated functions ($\min(X, k)$), or pairwise absolute differences ($|X_1 - X_2|$), always map out the new 'support' (image) first. Group the probabilities of underlying elementary events to avoid repetitive calculations.
The sample mean $\overline{X}$ is a standalone random variable whose variance is compressed by a factor of $\frac{1}{n}$. Recognizing cancellation between linear coefficients and sample sizes in $V(a\overline{X} + b) = a^2 \frac{V(X)}{n}$ is the key to solving advanced sampling problems.
2. Official Exam Deep Dive: Equispaced Periodic PMF & Parameter Inference
Step 1 (Compressing Probability Parameters): The support of $X$ is $\{0, 1, 2, 3, 4\}$. By the condition with step size 2:
$P(X = 0) = P(X = 2) = P(X = 4) = p$
$P(X = 1) = P(X = 3) = q$
Thus, five probabilities reduce to just two unknowns ($p$ and $q$).
Step 2 (Sum of Probabilities Condition):
$\sum_{x=0}^4 P(X = x) = 3p + 2q = 1 \implies 2q = 1 - 3p$
Step 3 (Evaluating Second Moment $E(X^2)$):
$E(X^2) = 0^2(p) + 1^2(q) + 2^2(p) + 3^2(q) + 4^2(p) = 20p + 10q$
Notice that $10q = 5(2q) = 5(1 - 3p) = 5 - 15p$. Substituting this in:
$E(X^2) = 20p + (5 - 15p) = 5p + 5$
$5p + 5 = \dfrac{35}{6} \implies 5p = \dfrac{5}{6} \implies p = \dfrac{1}{6}$
▶ Click to verify the full computation & final answer
Answer: $\dfrac{1}{6}$
Do not solve for $q = \frac{1-3p}{2}$ directly. Notice that $10q$ is an exact integer multiple of $2q$ ($10q = 5 \times 2q$). Grouping multiple terms algebraically prevents fraction arithmetic errors and saves valuable seconds under exam pressure.
3. Mastery Practice: 10 Multi-Layered Problems
[Problem 01] A discrete random variable $X$ takes values in $\{-2, -1, 0, 1, 2\}$ with PMF given by $P(X = x) = c(|x| + k)$, where $c, k > 0$. If the transformed variable $Y = |X|$ satisfies $V(Y) = \dfrac{2}{9}$, determine the exact value of $V(X)$.
▶ View Solution (Problem 01)
Step 2 (Distribution of $Y$): $Y \in \{0, 1, 2\}$ with $P(Y=0) = ck$, $P(Y=1) = 2c(1+k)$, and $P(Y=2) = 2c(2+k)$. Normalization yields $c(5k+6) = 1$. Solving $V(Y) = E(Y^2) - [E(Y)]^2 = \dfrac{2}{9}$ gives $k = 1$ and $c = \dfrac{1}{11}$.
Step 3: Since $X^2 = Y^2$, we have $V(X) = E(X^2) = E(Y^2) = \dfrac{10(1) + 18}{11} = \dfrac{28}{11}$.
Answer: $\dfrac{28}{11}$
Remember the identity $Y^2 = (|X|)^2 = X^2$. Whenever a symmetric variable is folded via an absolute value, its second moments are identical: $E(X^2) = E(Y^2)$.
[Problem 02] Two fair 6-sided dice are rolled. Let $X$ be the absolute difference between the two outcomes. If $Y = \min(X, 3)$, find the value of $36 \cdot V(Y)$.
▶ View Solution (Problem 02)
$P(Y = 0) = \dfrac{6}{36}$, $P(Y = 1) = \dfrac{10}{36}$, $P(Y = 2) = \dfrac{8}{36}$, and $P(Y = 3) = P(X \ge 3) = \dfrac{6 + 4 + 2}{36} = \dfrac{12}{36}$.
Step 2 (Moments): $E(Y) = \dfrac{62}{36} = \dfrac{31}{18}$, $E(Y^2) = \dfrac{150}{36} = \dfrac{25}{6}$.
Step 3: $V(Y) = \dfrac{25}{6} - \left(\dfrac{31}{18}\right)^2 = \dfrac{134}{324}$. Hence, $36 \cdot V(Y) = 36 \times \dfrac{134}{324} = \dfrac{134}{9}$.
Answer: $\dfrac{134}{9}$
For truncated variables like $\min(X, k)$, combine the upper tail into a single cumulative probability: $P(Y = k) = P(X \ge k)$.
[Problem 03] A population random variable $X$ has distribution $P(X=1) = \dfrac{1}{4}$, $P(X=2) = \dfrac{1}{2}$, and $P(X=3) = \dfrac{1}{4}$. A random sample of size $n$ is drawn, yielding sample mean $\overline{X}$. If $V(4\overline{X} + 1) = \dfrac{1}{2}$, determine the sample size $n$.
▶ View Solution (Problem 03)
Step 2: $V(4\overline{X} + 1) = 16 V(\overline{X}) = 16 \cdot \dfrac{V(X)}{n} = 16 \cdot \dfrac{1/2}{n} = \dfrac{8}{n}$.
Setting $\dfrac{8}{n} = \dfrac{1}{2}$ gives $n = 16$.
Answer: $16$
When the population mean is an integer, calculating variance as $E((X - \mu)^2)$ is significantly faster than using $E(X^2) - [E(X)]^2$.
[Problem 04] A bag contains two balls labeled 1, one ball labeled 2, and one ball labeled 3. Two balls are selected simultaneously without replacement. Let $X$ be the product of the numbers on the two balls. Compute $V(3X - 2)$.
▶ View Solution (Problem 04)
Step 2: $E(X) = \dfrac{17}{6}$, $E(X^2) = \dfrac{63}{6} = \dfrac{21}{2}$.
Step 3: $V(X) = \dfrac{21}{2} - \left(\dfrac{17}{6}\right)^2 = \dfrac{89}{36}$. Thus, $V(3X - 2) = 9 V(X) = 9 \cdot \dfrac{89}{36} = \dfrac{89}{4}$.
Answer: $\dfrac{89}{4}$
Anticipate the algebraic cancellation: since $V(3X - 2) = 9V(X)$ and the denominator of $V(X)$ is $6^2 = 36$, the 9 and 36 will cancel to leave a denominator of 4.
[Problem 05] A random variable $X$ follows a binomial distribution $B(n, p)$ such that $E(X) = 15$ and $P(X = 1) = 30 \cdot P(X = 0)$. From this population, a random sample of size $25$ is drawn to obtain sample mean $\overline{X}$. Find $V(5\overline{X})$.
▶ View Solution (Problem 05)
Step 2: Population variance $V(X) = np(1-p) = 30(1/2)(1/2) = \dfrac{15}{2}$.
Step 3: $V(5\overline{X}) = 25 V(\overline{X}) = 25 \cdot \dfrac{V(X)}{25} = V(X) = \dfrac{15}{2}$.
Answer: $\dfrac{15}{2}$
Notice that $V(\sqrt{N}\overline{X}) = N \frac{\sigma^2}{N} = \sigma^2$. Whenever the linear coefficient squared matches the sample size, the sample variance collapses directly back to the population variance.
[Problem 06] A discrete random variable $X$ has PMF $P(X = x) = \dfrac{k}{x(x+1)}$ for $x \in \{1, 2, 3, 4, 5\}$. If $Y = \dfrac{1}{X+1}$, find $E(Y)$.
▶ View Solution (Problem 06)
Step 2: $E(Y) = \sum_{x=1}^5 \dfrac{1}{x+1} \cdot \dfrac{k}{x(x+1)} = \dfrac{6}{5} \sum_{x=1}^5 \dfrac{1}{x(x+1)^2}$. Evaluating each term gives $E(Y) = \dfrac{23}{54}$.
Answer: $\dfrac{23}{54}$
Telescoping series frequently appear in PMF normalization. Normalize $k$ first, then transform the variable inside the summation.
[Problem 07] Two discrete random variables $X$ and $Y$ each take values in $\{1, 2, 3, 4\}$. For each $k \in \{1, 2, 3, 4\}$, $P(Y = k) = \dfrac{1}{2} P(X = k) + \dfrac{1}{8}$. Given that $V(X) = 1$ and $X$ is symmetric about $2.5$, find $V(Y)$.
▶ View Solution (Problem 07)
Step 2: $E(U) = \dfrac{5}{2}$, $E(U^2) = \dfrac{15}{2}$. Since $E(X) = \dfrac{5}{2}$, we have $E(Y) = \dfrac{5}{2}$.
Step 3: $E(X^2) = V(X) + [E(X)]^2 = 1 + \dfrac{25}{4} = \dfrac{29}{4}$.
$E(Y^2) = \dfrac{1}{2}\left(\dfrac{29}{4}\right) + \dfrac{1}{2}\left(\dfrac{30}{4}\right) = \dfrac{59}{8}$.
$V(Y) = \dfrac{59}{8} - \left(\dfrac{5}{2}\right)^2 = \dfrac{9}{8}$.
Answer: $\dfrac{9}{8}$
Use the linearity of moments on distribution mixtures: $E(Y^r) = \alpha E(X^r) + (1-\alpha) E(U^r)$.
[Problem 08] A non-negative discrete random variable $X$ takes values in $\{0, 1, 2, 3\}$ with survival function $P(X \ge k) = \left(\dfrac{3-k}{3}\right)^2$ for $k \in \{0, 1, 2, 3\}$. Find $18 \cdot E(X)$.
▶ View Solution (Problem 08)
Step 2: $P(X \ge 1) = \dfrac{4}{9}$, $P(X \ge 2) = \dfrac{1}{9}$, $P(X \ge 3) = 0$.
$E(X) = \dfrac{4}{9} + \dfrac{1}{9} + 0 = \dfrac{5}{9} \implies 18 \cdot E(X) = 18 \times \dfrac{5}{9} = 10$.
Answer: $10$
The tail-sum formula $E(X) = \sum_{k=1}^\infty P(X \ge k)$ avoids computing individual probabilities $P(X=k)$ entirely.
[Problem 09] A bag contains 3 white balls and 2 black balls. Balls are drawn one by one without replacement until the first white ball is drawn. Let $X$ be the total number of balls drawn. Compute $V(X)$.
▶ View Solution (Problem 09)
$P(X=1) = \dfrac{3}{5} = \dfrac{6}{10}$, $P(X=2) = \dfrac{2}{5} \times \dfrac{3}{4} = \dfrac{3}{10}$, $P(X=3) = \dfrac{2}{5} \times \dfrac{1}{4} \times 1 = \dfrac{1}{10}$.
Step 2: $E(X) = \dfrac{3}{2}$, $E(X^2) = \dfrac{27}{10}$.
$V(X) = \dfrac{27}{10} - \dfrac{9}{4} = \dfrac{9}{20}$.
Answer: $\dfrac{9}{20}$
In sampling without replacement, the maximum number of trials until the first success is always bounded by $(\text{number of failures} + 1)$.
[Problem 10] Two independent random samples of sizes $n_1 = 4$ and $n_2 = 9$ are drawn from a population with variance $\sigma^2$, producing sample means $\overline{X}_1$ and $\overline{X}_2$. Consider the unbiased estimator $W = \alpha \overline{X}_1 + (1-\alpha) \overline{X}_2$. Determine the optimal weight $\alpha$ that minimizes $V(W)$, and express this minimum variance in terms of $\sigma^2$.
▶ View Solution (Problem 10)
Step 2: Minimizing the quadratic function $f(\alpha) = \dfrac{13\alpha^2 - 8\alpha + 4}{36}$ yields vertex $\alpha = \dfrac{4}{13}$.
Step 3: Minimum variance is $f(4/13) \sigma^2 = \dfrac{1}{13} \sigma^2$.
Answer: $\alpha = \dfrac{4}{13}$, Minimum Variance $= \dfrac{1}{13}\sigma^2$
Inverse-variance weighting: the optimal weights for combining independent estimators are proportional to their precisions (sample sizes): $\alpha : (1-\alpha) = n_1 : n_2 = 4 : 9 \implies \alpha = \frac{4}{13}$.
3 Killer Challenge Problems: Structural Inference
Moving beyond routine calculations: these problems integrate 2D lattice metrics, non-linear functional moments, and Markovian state-transition mechanics.
[Killer Challenge 01] A discrete random variable $X$ has support $\{1, 2, 3, 4, 5\}$ and satisfies $P(X = 3-k) = P(X = 3+k)$ for $k \in \{1, 2\}$. A real-valued function $g(t) = E\left((t - X)^2\right)$ attains its minimum value of $\dfrac{5}{2}$ at $t = 3$. For the transformed variable $Y = (X-3)^3$, compute $V(Y)$.
▶ View Solution (Killer 01)
$g(t) = t^2 - 2tE(X) + E(X^2) = (t - E(X))^2 + V(X)$.
The vertex $(3, 5/2)$ directly implies $E(X) = 3$ and $V(X) = \dfrac{5}{2}$.
Step 2 (Odd Transformation & Mean Zero):
Let $Z = X - 3 \in \{-2, -1, 0, 1, 2\}$, which is symmetric about 0.
Since $Y = Z^3$ is an odd function of $Z$, $Y$ is also symmetric about 0. Hence, $E(Y) = 0$ and $V(Y) = E(Y^2) = E(Z^6)$.
Step 3 (Solving Higher-Order Moments):
Let $P(Z=0) = p_0$, $P(|Z|=1) = 2p_1$, $P(|Z|=2) = 2p_2$.
$V(X) = E(Z^2) = 2p_1 + 8p_2 = \dfrac{5}{2} \implies p_1 + 4p_2 = \dfrac{5}{4}$.
$E(Y^2) = E(Z^6) = 2(1^6)p_1 + 2(2^6)p_2 = 2p_1 + 128p_2 = 2(p_1 + 4p_2) + 120p_2$.
With boundary constraints, $V(Y) = \dfrac{5}{2} + 120\left(\dfrac{1}{4}\right) = 32$.
Answer: $32$
$E((t-X)^2)$ is always minimized at $t = \mu$ with minimum value $\sigma^2$. Any odd power of a centered symmetric variable has expectation zero, immediately reducing variance to a pure second-moment calculation.
[Killer Challenge 02] Consider the 9 lattice points $(x, y)$ where $x, y \in \{0, 1, 2\}$. Two distinct points are chosen simultaneously at random. Let $X$ be the Euclidean distance between them. For $Y = X^2$, compute $18 \cdot V(Y)$.
▶ View Solution (Killer 02)
Total pairs: $\binom{9}{2} = 36$.
• $Y = 1$: 12 pairs ($P = 12/36$)
• $Y = 2$: 8 pairs ($P = 8/36$)
• $Y = 4$: 6 pairs ($P = 6/36$)
• $Y = 5$: 8 pairs ($P = 8/36$)
• $Y = 8$: 2 pairs ($P = 2/36$)
Step 2 (Expectation and Variance via Deviations):
$E(Y) = \dfrac{1(12) + 2(8) + 4(6) + 5(8) + 8(2)}{36} = \dfrac{108}{36} = 3$.
Since $E(Y)$ is an exact integer, evaluate $V(Y) = E((Y-3)^2)$:
$V(Y) = \dfrac{(-2)^2(12) + (-1)^2(8) + (1)^2(6) + (2)^2(8) + (5)^2(2)}{36} = \dfrac{144}{36} = 4$.
Step 3: $18 \cdot V(Y) = 18 \times 4 = 72$.
Answer: $72$
Do not compute with irrational distances ($X \in \{1, \sqrt{2}, 2, \sqrt{5}, 2\sqrt{2}\}$). Upgrade the squared distance $Y = X^2$ into your primary variable to keep all arithmetic cleanly in the integer domain.
[Killer Challenge 03] A particle starts at the origin $0$ on a number line. A fair coin is tossed repeatedly. If Heads appears, the particle moves $+1$; if Tails appears, it remains in place. However, immediately after reaching any odd position, the particle is forced to move $-1$ on the subsequent step regardless of the coin toss. After exactly 3 coin tosses, let $X$ be the final position. In an independent random sample of 100 trials, let $\overline{X}$ be the sample mean. Compute $1000 \cdot V(\overline{X})$.
▶ View Solution (Killer 03)
The particle can only end at position $0$ or $1$.
Paths ending at $X = 1$:
1) $[H \to \text{Forced Reset} \to H]$: $\dfrac{1}{2} \times 1 \times \dfrac{1}{2} = \dfrac{1}{4}$
2) $[T \to T \to H]$: $\dfrac{1}{2} \times \dfrac{1}{2} \times \dfrac{1}{2} = \dfrac{1}{8}$
$\implies P(X = 1) = \dfrac{1}{4} + \dfrac{1}{8} = \dfrac{3}{8}$, and $P(X = 0) = \dfrac{5}{8}$.
Step 2 (Bernoulli Variance & Sampling):
Since $X \in \{0, 1\}$, $X \sim \text{Bernoulli}(3/8)$.
$V(X) = p(1-p) = \dfrac{3}{8} \times \dfrac{5}{8} = \dfrac{15}{64}$.
For $n = 100$: $V(\overline{X}) = \dfrac{15/64}{100} = \dfrac{15}{6400}$.
$1000 \cdot V(\overline{X}) = 1000 \times \dfrac{15}{6400} = \dfrac{75}{32}$.
Answer: $\dfrac{75}{32}$
When an intricate stochastic process collapses into a binary outcome $\{0, 1\}$, instantly apply the Bernoulli variance shortcut $V(X) = p(1-p)$.

Comments
Post a Comment