A random variable is not a variable in the algebraic sense — it is a function. Specifically, it maps each outcome of a random experiment to a real number. When you roll a die, the outcome "face shows 4" becomes the number 4. That mapping is a random variable.
Think of it this way: a random experiment produces messy outcomes (heads/tails, defective/non-defective, 1 through 6). A random variable assigns a clean numerical label to each outcome so you can do arithmetic on it — calculate averages, spreads, and probabilities. Without this translation, you cannot define "expected profit" or "average waiting time."
There are two types:
Discrete random variables take countable values — often whole numbers. The number of defects in a batch, the result of rolling a die, the number of correct answers in a quiz. You can list every possible value.
Continuous random variables take any value within an interval. Height, weight, time until failure, temperature. Between any two values, there are infinitely many possibilities. You cannot list them all — instead, you describe them using a curve (the probability density function, or PDF).
The analogy that works: think of a discrete distribution as a bar chart — each bar has a specific height (probability). A continuous distribution is a smooth curve — probabilities correspond to areas under the curve, not heights at points. The probability that a continuous variable equals exactly one value is zero — only intervals have non-zero probability.
What SSC CGL actually tests here is focused: mean (expected value), variance, and the three named distributions — Binomial, Poisson, and Uniform. If you learn nothing else from this chapter, learn the six formulas in the table below cold. Every PYQ in this topic traces back to one of those six.
For a discrete random variable X with values x₁, x₂, ..., xₙ and probabilities P(X = xᵢ):
This is a weighted average — each value weighted by its probability. You are not averaging the values themselves; you are averaging them according to how likely they are.
For a continuous random variable:
where f(x) is the PDF. SSC CGL rarely asks you to evaluate this integral directly — they give you named distributions with known formulas.
Variance measures spread around the mean:
Or equivalently, E[(X - μ)²]. The standard deviation is σ = √Var(X).
A critical shortcut: for named distributions, use the canned formulas below — do not try to compute variance from first principles in the exam.
Setup: n independent trials, each with probability p of success, q = 1 − p of failure.
| Parameter | Formula |
|-----------|---------|
| Mean | μ = np |
| Variance | σ² = npq = np(1−p) |
| Standard Deviation | σ = √(npq) |
The PMF (probability mass function): P(X = k) = C(n,k) · pᵏ · qⁿ⁻ᵏ
Look — the variance npq is always less than the mean np (since q < 1). This is your internal check. If your calculated variance exceeds the mean for a Binomial, something is wrong.
Common distractor trap: When n = 10, p = 0.4, the mean is 4 and the variance is 2.4. Examiners put both in the options. The variance npq = 10 × 0.4 × 0.6 = 2.4 — not 4 (mean), not 6 (range filler).
Setup: counts the number of rare events occurring in a fixed interval of time or space. λ is the average rate.
| Parameter | Formula |
|-----------|---------|
| Mean | μ = λ |
| Variance | σ² = λ |
This is the single most important fact about the Poisson distribution: mean equals variance equals λ. No other standard distribution has this property. The standard deviation is √λ.
The PMF: P(X = k) = e⁻λ · λᵏ / k!
SSC CGL will test only the mean-variance equality, not the PMF computation.
Setup: every value in [a, b] is equally likely. The PDF is the flat line f(x) = 1/(b−a).
| Parameter | Formula |
|-----------|---------|
| Mean | μ = (a + b) / 2 |
| Variance | σ² = (b − a)² / 12 |
The mean is simply the midpoint. The variance formula with 12 in the denominator is the one people forget — commit it precisely as (b−a)²/12.
For U(2, 8): mean = (2+8)/2 = 5, variance = (8−2)²/12 = 36/12 = 3. Straightforward arithmetic, but the distractor is usually the range (6) or range² (36).
The CDF F(x) = P(X ≤ x) gives the probability that X takes a value at most x.
For discrete distributions: F(x) = Σ P(X = xᵢ) for all xᵢ ≤ x
For continuous distributions: F(x) = ∫₋∞ˣ f(t) dt
Key properties every aspirant should know:
F(−∞) = 0, F(+∞) = 1f(x) = dF(x)/dxSSC CGL tests CDF primarily through recognition questions — identifying which function is a valid CDF, or reading off probabilities from a given CDF.
The bell curve. SSC CGL tests only:
Z = (X − μ) / σ transforms any normal to N(0,1)You are not expected to integrate the normal PDF in SSC CGL. Questions are conceptual or use Z-table look-ups.
For Binomial B(n, p): variance = np × q, which is np multiplied by something less than 1. So variance is always less than mean. If you compute variance > mean for a Binomial, you have made an error — use this as a self-check. Calculation check: B(10, 0.4) — mean = 4, variance = 4 × 0.6 = 2.4. One multiplication after you have the mean: standard method needs re-deriving npq (3 multiplications), this shortcut uses mean × q (1 multiplication, ~8s vs ~20s).
For a Poisson variable, mean = variance = λ. If a question gives you λ and asks for variance, write it down immediately — no calculation needed. Conversely, if it gives variance and asks for mean, same answer. This collapses a potential 30-second computation into a 3-second read. Watch out: standard deviation = √λ, not λ. Examiners insert √λ and λ² as distractors specifically because students confuse variance with standard deviation. Pattern recognition eliminates 2 wrong options instantly.
For U(a, b), mean = (a + b)/2. You do not need any formula — it is the midpoint of the interval. For variance, the only formula to memorize is (b−a)²/12. Notice the denominator is 12, not 6 or 4. Verify once: U(0,1) has variance 1/12 ≈ 0.083, which makes sense since the values are all between 0 and 1. Uniform mean: 2 seconds (mental midpoint). Uniform variance: write (b−a), square it, divide by 12. Three arithmetic steps vs attempting derivation from scratch (~60s).
Write probabilities and values as two columns, multiply across each row, then add the products. For X = {1, 2, 3} with P = {0.2, 0.5, 0.3}: compute 0.2, 1.0, 0.9 — then sum to 2.1. Avoid working left-to-right in your head, which risks accumulation errors. The column method has 3 multiplications + 2 additions done sequentially. Left-to-right mental accumulation has the same count but higher error rate under time pressure. Column method: ~15s vs mental running sum: ~25s with higher error rate.
For Binomial B(n, p) questions, examiners always include the mean (np) as a distractor for variance, and np(1−p) for mean. Before computing anything, label your answer choices: "this looks like np," "this looks like npq," "this looks like √npq." Then compute the one you actually need. Eliminates 2 of 4 options before any calculation. For B(10, 0.4): options 4.0 (=np, mean) and 2.4 (=npq, variance) will both appear — knowing which is which before calculating saves one full computation cycle (~15s saved).
When you see a random variable / distribution question in the exam, run this decision tree:
Step 1 — Identify the distribution. Does the question say "Binomial," "Poisson," or "Uniform"? If yes, go directly to the formula table in your head.
Step 2 — Identify what is being asked. Mean, variance, or standard deviation?
Step 3 — Apply the single correct formula:
np | Binomial variance: np(1−p)λ(a+b)/2 | Uniform variance: (b−a)²/12Step 4 — Check your answer against the options. If computing variance for Binomial, verify it is less than the mean. If computing Poisson variance, verify it equals the mean you would compute.
Step 5 — For discrete E(X) questions: multiply each value by its probability, sum. No named distribution needed.
If none of this applies and the question gives a PDF or CDF with integrals, it is likely beyond SSC CGL scope — but check whether the question is asking for a property (like total area = 1) rather than a computation.
Total time target: 45–60 seconds per question in this category.
Why this question: This is the most direct test of the Binomial mean formula — pure recall with a distractor set designed to catch formula confusion.
Solving path: Binomial B(10, 0.5). Mean = n × p = 10 × 0.5 = 5. Options 0.5 (just p), 2.5 (np/2), and 10 (just n) are all designed to trap students who half-remember the formula. The answer 5 is straightforward. Time: 10 seconds.
Why this question: Tests discrete E(X) computation — the most fundamental operation with random variables.
Solving path: E(X) = (1 × 0.2) + (2 × 0.5) + (3 × 0.3). Column multiply: 0.2, 1.0, 0.9. Sum: 2.1. The distractor 2.0 is close — verify by noting that since P(X=3) = 0.3, the distribution is slightly right-weighted, pulling E(X) above 2.0. Answer: 2.1. Time: 20 seconds.
Why this question: Tests whether you know that variance ≠ mean for Binomial, and that both values appear as distractors in the option set.
Solving path: B(10, 0.4). The mean is 10 × 0.4 = 4.0 — this appears as option (b), the trap. The variance is np(1−p) = 10 × 0.4 × 0.6 = 2.4. Answer: 2.4. Notice that 1.6 = 10 × 0.4 × 0.4 (wrong — uses p² instead of pq). Eliminate it immediately. Time: 25 seconds.
Why this question: Tests the unique property of Poisson distribution — mean equals variance. The distractors are λ², √λ, and 2λ, covering the most common confusions.
Solving path: Poisson with λ = 3. Mean = λ = 3. Variance = λ = 3. Standard deviation = √3 ≈ 1.73 (distractor). λ² = 9 (distractor). The question asks for variance. Answer: 3. Time: 8 seconds — pure recall.
Why this question: Tests the Uniform distribution variance formula, which is the most commonly forgotten formula in this chapter.
Solving path: U(2, 8). Variance = (b−a)²/12 = (8−2)²/12 = 36/12 = 3. The range (b−a) = 6 appears as a distractor. Range² = 36 appears as a distractor. The correct answer 3 requires remembering the /12 step. Mean = (2+8)/2 = 5, also in options for a related question — not the same as variance. Answer: 3. Time: 20 seconds.
Confusing Binomial mean and variance. When B(n, p) options include both np and np(1−p), students pick the larger value thinking "variance is bigger." Variance is always smaller than mean for Binomial. The mean is np; variance is np(1−p).
Poisson standard deviation vs variance. The variance is λ. The standard deviation is √λ. Examiners put √λ in the options when asking for variance. If λ = 4, variance = 4, standard deviation = 2. These are never the same unless λ = 1.
Uniform variance denominator. Students write (b−a)²/6 or (b−a)/12. The correct formula is (b−a)²/12 — square first, then divide by 12. No exceptions.
Using the wrong q in Binomial variance. Variance = np(1−p), not np². The (1−p) factor is the probability of failure, q. Using p² instead of pq is one of the most frequent arithmetic errors — it gives a smaller variance that will still appear in the option set.
Treating continuous distributions like discrete for probabilities at a point. For a continuous random variable, P(X = 5) = 0 exactly. Only P(a ≤ X ≤ b) has non-zero value. If a question asks for the probability at a single point for a continuous distribution, the answer is 0 — not f(5).
Adding instead of multiplying in E(X). E(X) = Σ x·P(X=x), not Σ x + Σ P(X=x). The sum of probabilities always equals 1 — if you add rather than multiply, you will get a nonsense answer that may still look plausible among the options.