Statistics › Section D · Discrete random variables › Topic D.3
The binomial distribution
Flip, check, guess - whenever a fixed number of independent yes-or-no trials produces a count of successes, the binomial distribution gives the exact probabilities of every possible count. It is the workhorse of discrete probability.
Counting successes
Flip a coin 10 times and count heads. Check 20 items and count defectives. Guess on 10 questions and count correct answers. Each story has the same skeleton: a set number of trials, each a success or a failure, and a total to report. Write \(X \sim B(n, p)\) for such a count, read "\(X\) is binomial with \(n\) trials and success chance \(p\)". Quality control, polling, clinical trials and multiple-choice tests all lean on it.
Binomial settings: check BINS
Four conditions make a setting binomial. Binary: each trial ends in success or failure. Independent: trials do not influence each other. Number: the count of trials, \(n\), is fixed in advance. Same probability: every trial succeeds with the same chance \(p\). Rolling a die 8 times and counting sixes fits. Drawing cards without replacement from a small deck does not - each draw changes \(p\). The exception worth knowing: drawing from a very large batch keeps \(p\) nearly constant, so quality sampling from a big production run is treated as binomial.
The binomial formula
The probability of exactly \(k\) successes in \(n\) trials is
The formula splits the job in two. Any one specific arrangement - say successes on trials 1, 4 and 7 - has probability \(p^k(1-p)^{n-k}\): each success contributes \(p\), each failure \(1-p\). And the combination \(\binom{n}{k}\) counts how many distinct arrangements hold exactly \(k\) successes. One arrangement's probability, times the number of arrangements, gives the total.
Mean and standard deviation
A binomial count centres and spreads by formulas that skip the table entirely: \(\mu = np\) and \(\sigma = \sqrt{np(1-p)}\), the root of the variance \(\sigma^2 = np(1-p)\). In 100 fair flips, expect \(100 \times 0.5 = 50\) heads, with \(\sigma = \sqrt{100 \times 0.5 \times 0.5} = \sqrt{25} = 5\). In 20 flips, the mean is 10 and \(\sigma = \sqrt{5} \approx 2.24\). The spread term \(p(1-p)\) is largest at \(p = 0.5\): coin flips are the most unpredictable per trial, while extreme \(p\) makes outcomes routine.
Worked example: exactly three heads in five flips
Count heads in \(n = 5\) fair flips, so \(p = 0.5\), aiming for exactly \(k = 3\). 1. One arrangement - say H H H T T - has probability \((0.5)^3(0.5)^2 = (0.5)^5 = \frac{1}{32}\). 2. The number of arrangements is \(\binom{5}{3} = 10\). 3. \(P(X = 3) = 10 \times \frac{1}{32} = \frac{10}{32} = 0.3125\). 4. Check: the full distribution runs \(\frac{1}{32}, \frac{5}{32}, \frac{10}{32}, \frac{10}{32}, \frac{5}{32}, \frac{1}{32}\) for \(k = 0\) to 5, totalling \(\frac{32}{32} = 1\) - and 3 heads mirrors 2 heads, both \(\frac{10}{32}\).
Worked example: defectives in a sample
A large batch of items has a 10% defective rate, so \(p = 0.1\); an inspector samples \(n = 8\). a) Probability of exactly 2 defectives: \(\binom{8}{2}(0.1)^2(0.9)^6 = 28 \times 0.01 \times 0.531441 \approx 0.149\). b) Probability of at least one defective: the complement of a clean sample, \(1 - (0.9)^8 = 1 - 0.4305 \approx 0.570\). Check: (b) is not \(8 \times 0.1\) - that is the expected count, not a probability - and 0.570 exceeds 0.149 sensibly, since at least one covers many more cases than exactly two.
"At least" questions
Binomial questions rarely stop at exactly. "At least \(k\)" adds \(P(k), P(k+1), \ldots, P(n)\); "at most \(k\)" adds the terms from 0 to \(k\). The complement shortcut does heavy lifting: at least one success is \(1 - P(0)\), one term instead of many. A student guesses all 10 questions of a four-choice quiz: \(p = 0.25\), expected score \(np = 2.5\), and \(P(\text{at least 6 correct})\) sums five terms to about 0.02. Guessing is a poor strategy, and the binomial model shows exactly how poor: about one chance in fifty of reaching 6.
Shape of binomial distributions
The binomial histogram's shape follows \(p\). At \(p = 0.5\) it is perfectly symmetric - the five-flip distribution mirrors around 2.5. Small \(p\) piles successes at the low end with a long right tail: skewed right. Large \(p\) piles them high with a left tail: skewed left. As \(n\) grows, the histogram turns bell-shaped, and the rule of thumb \(np \ge 10\) and \(n(1-p) \ge 10\) marks where a normal curve approximates it well - a connection the normal-distribution chapter develops fully.
Common mistakes to avoid
First, applying the formula without checking BINS - a changing \(p\) or dependent trials void it. Second, forgetting the combination factor \(\binom{n}{k}\): one arrangement's probability is not the whole answer. Third, confusing \(np\), an expected count, with a probability, or using \(np\) as the SD instead of \(\sqrt{np(1-p)}\). Fourth, reading "at least one" as \(P(X = 1)\): it spans every case except zero. Fifth, sampling without replacement from a small population and calling it binomial. Sixth, reading "at most 2" as \(P(X < 2)\): it includes \(X = 2\), so add \(P(0) + P(1) + P(2)\).
Key takeaways
Binomial counts successes in \(n\) independent yes-or-no trials with constant success chance \(p\) - check BINS first. \(P(X = k) = \binom{n}{k}p^k(1-p)^{n-k}\): the number of arrangements times one arrangement's probability. The centre is \(\mu = np\), the spread \(\sigma = \sqrt{np(1-p)}\). "At least" and "at most" questions add table terms, with \(1 - P(0)\) the fast route to at least one. Shape runs symmetric at \(p = 0.5\), skewed for extreme \(p\), and bell-shaped once \(n\) is large enough that \(np\) and \(n(1-p)\) both reach about 10.
Flashcards
Flashcard
A count of successes in fixed, independent yes-or-no trials
Test your understanding
Question 1 of 5
Which scenario is a binomial setting?
Question 2 of 5
In 5 fair coin flips, what is the probability of exactly 3 heads?
Question 3 of 5
A fair coin is flipped 20 times. What are the expected number of heads and its standard deviation?
Question 4 of 5
A batch has a 10% defective rate and 8 items are sampled. What is the probability of at least one defective?
Question 5 of 5
A binomial variable has n = 8 trials and p = 0.1. What is its expected number of successes?
