top of page

ISS 2016 Statistics Paper-1 Solution: Question 46 (Skewness of the Binomial Distribution)

Oct 7
6 min read

Continuing our question-by-question walk through the ISS Statistics Paper-1 archive, today we pick up right where the last post left off and tackle Question 46 from the 2016 paper. This one sits in the probability distributions section and asks you to work out a skewness formula for the Binomial distribution from first principles — a question that rewards a careful, methodical approach far more than memorisation.


Quick Summary


  • Topic:Binomial Distribution — Coefficient of Skewness

  • Question Reference:ISS 2016, Statistics Paper-1, Question 46

  • Correct Answer:(b) (1 − 2p) / √(npq)

The Question, As Asked

The probability of a random variable X is given by:


P(X = r) = nCr pr qn−r; r = 0, 1, 2, ..., n; p + q = 1, p > 0.


What is the coefficient of skewness of the random variable X?


  • (a) (1 + 2p) / √(npq)

  • (b) (1 − 2p) / √(npq)

  • (c) −p / √(nq)

  • (d) (1 + p) / √(nq)

Step 1: Recognise the Distribution

The pmf P(X = r) = nCr pr qn−r is exactly the probability mass function of a Binomial(n, p)random variable — X counts the number of "successes" in n independent trials, each with success probability p and failure probability q = 1 − p. The coefficient of skewness (the standardised third moment, usually written γ1 = μ3/σ3) tells us whether the distribution leans to the left or the right of its mean. To get there cleanly, instead of summing r3 P(X = r) directly — which gets messy fast — we build the moment generating function (MGF) and extract everything we need from it, step by step.

Step 2: Write Down the MGF of X

By definition, MX(t) = E[etX] = Σr=0n nCr pr qn−r etr.


Group etr with pr as (p et)r:


MX(t) = Σr=0n nCr (p et)r qn−r


This sum is exactly the binomial expansion of (q + p et)n. So:


MX(t) = (q + p et)n

Step 3: Convert to the Cumulant Generating Function

Raw moments from an MGF require differentiating a power of an exponential, which gets tedious by the third derivative. Cumulants are far friendlier for skewness because the third cumulant κ3 equals the third central moment μ3 directly (no correction terms needed, unlike for κ4). So we take logs:


K(t) = ln MX(t) = n ln(q + p et)

Step 4: Expand K(t) as a Power Series in t

First expand et using its Taylor series et = 1 + t + t2/2 + t3/6 + ... (ignoring t4 and higher, since we only need up to the t3 term):


q + p et = q + p(1 + t + t2/2 + t3/6) = (p + q) + pt + (p/2)t2 + (p/6)t3


Since p + q = 1, this simplifies to 1 + x, where x = pt + (p/2)t2 + (p/6)t3.


Now use ln(1 + x) = x − x2/2 + x3/3 − ... and substitute, keeping only terms up to t3:


x2 = p2t2 + p2t3 + (higher order terms)


x3 = p3t3 + (higher order terms)


So:


ln(1 + x) = [pt + (p/2)t2 + (p/6)t3] − (1/2)[p2t2 + p2t3] + (1/3)[p3t3] + ...


Collecting by power of t:


  • Coefficient of t1: p

  • Coefficient of t2: p/2 − p2/2

  • Coefficient of t3: p/6 − p2/2 + p3/3


Multiplying the whole expansion by n gives K(t) = n[pt + (p/2 − p2/2)t2 + (p/6 − p2/2 + p3/3)t3 + ...].


Pro Tip: This is the exact step where most students lose marks under time pressure — forgetting that the cumulant generating function is defined as K(t) = Σ κr tr/r!, not Σ κr tr. You must multiply the coefficient of tr in the expansion by r! to recover κr. Skipping this factorial step is the single most common reason students land on option (a) instead of (b) — the sign comes out right, but the magnitude is off. The way this derivation is usually taught is to write out "coefficient × r! = κr" as a standing rule before touching any specific distribution, so the habit is automatic by the time you meet Binomial, Poisson, or Negative Binomial skewness questions in the exam.

Step 5: Extract the Cumulants

Using κr = r! × (coefficient of tr in K(t)):


κ1 = n × p × 1! = np   (this is the mean — matches the familiar E[X] = np)


κ2 = n × (p/2 − p2/2) × 2! = n(p − p2) = np(1 − p) = npq   (this is the variance — matches Var(X) = npq)


κ3 = n × (p/6 − p2/2 + p3/3) × 3! = n(p − 3p2 + 2p3)


Factor the cubic: p − 3p2 + 2p3 = p(1 − 3p + 2p2) = p(1 − p)(1 − 2p). Since q = 1 − p:


κ3 = n p q (1 − 2p) = μ3 (the third central moment, since κ3 = μ3 always)

Step 6: Compute the Coefficient of Skewness

The (moment) coefficient of skewness is γ1 = μ3 / σ3, where σ3 = (npq)3/2 since σ2 = npq.


γ1 = n p q (1 − 2p) / (npq)3/2


Cancelling one factor of npq from numerator and denominator:


γ1 = (1 − 2p) / √(npq)


This is exactly option (b), and it matches the official ISS 2016 answer key for Question 46 (Series A), which also records (b) as correct. Our independent derivation and the key are in full agreement.

Why This Question Matters

Skewness formulas for named discrete distributions (Binomial, Poisson, Negative Binomial) are a recurring favourite in ISS Paper-1 because they test three things at once: whether you can correctly identify a distribution from its pmf, whether you can generate and manipulate an MGF or CGF confidently, and whether you can keep track of algebraic sign conventions under time pressure. The underlying result — that Binomial skewness is positive when p < 0.5, negative when p > 0.5, and exactly zero when p = 0.5 — is also genuinely useful intuition: it tells you, without any calculation, which way a binomial histogram will lean just by looking at whether success is "rare" or "common." Students who drill the MGF-to-cumulant pipeline until it becomes automatic rarely get tripped up by the algebra when a slightly different distribution (say, Poisson or Geometric) shows up in its place on exam day.

Frequently Asked Questions

What is the general formula for the coefficient of skewness of a Binomial(n, p) distribution?

It is γ1 = (1 − 2p) / √(npq), which can equivalently be written as (q − p) / √(npq) since q = 1 − p. Both forms appear interchangeably in textbooks and answer keys, so recognise them as the same expression.

Why does the Binomial distribution become positively skewed when p is small?

When p < 0.5, successes are relatively rare, so most of the probability mass sits near the lower values of X with a long tail stretching toward higher values — that long right tail is what positive skewness means. Plugging p < 0.5 into (1 − 2p)/√(npq) correctly gives a positive number, confirming this intuition algebraically.

What is the difference between the moment coefficient of skewness and Pearson's coefficient of skewness?

The moment coefficient, γ1 = μ3/σ3, is based on the third central moment and is the standard definition used in MGF/CGF-based derivations like this one. Pearson's coefficient, by contrast, is an approximate measure defined as 3(mean − median)/standard deviation, and the two need not give identical numerical values for every distribution.

Do I need to memorise the cumulants of the Binomial distribution, or should I derive them each time?

For an exam like ISS, it is worth memorising κ1 = np, κ2 = npq, and κ3 = npq(1 − 2p) as standard results, but you should still be comfortable deriving them from the MGF under exam pressure, since a slightly modified question (for example, asking for kurtosis, or for Y = X − 2 as seen in a neighbouring question on the same paper) can only be solved correctly if you understand where the formulas come from.

Is the Binomial distribution ever perfectly symmetric?

Yes — when p = 0.5, (1 − 2p) becomes zero, so γ1 = 0 and the distribution is exactly symmetric about its mean np. This is easy to verify directly too: with p = q = 0.5, P(X = r) and P(X = n − r) become equal for every r, which is the definition of symmetry here.

Why is the cumulant generating function preferred over direct summation for finding higher moments?

Direct summation of r3 P(X = r) requires handling awkward combinatorial identities and rarely simplifies cleanly. The CGF approach instead turns the problem into straightforward algebra — a Taylor expansion and term collection — and the relationship κ3 = μ3 means you get the third central moment with no extra correction terms, unlike when converting raw moments to central moments directly.


That wraps up Question 46 of the ISS 2016 Statistics Paper-1 series. If any step in the cumulant expansion felt unclear, or you'd like to see the same approach applied to Poisson or Negative Binomial skewness, drop a comment below — and if this helped, do share it with a fellow ISS aspirant who's working through the same paper.

Recent Posts

See All

Comments


bottom of page