Content
Summary
The binomial distribution models the count of successes in n independent Bernoulli trials with constant probability p. It is defined by the probability mass function Pr(X=k) = (n choose k) p^k q^(n-k), where q = 1-p. Key properties include a mean of np, variance of npq, and symmetry when p=0.5.
Intuition
Think of a binomial experiment as a sequence of yes/no decisions, like checking if cells are neutrophils. Each trial is independent, meaning one outcome doesn't affect the next. The 'success' label is neutral; it just marks the event we are counting, which could be a disease or a default. The total count of successes follows a specific pattern determined by how many ways you can arrange those successes among the trials, captured by the binomial coefficient.
Key equations
The binomial random variable X is the sum of n independent Bernoulli trials.
The probability mass function for exactly k successes in n trials.
The binomial coefficient counting distinct arrangements of k successes.
The expected number of successes (mean).
The variance of the distribution.
Derivation
- Define a single Bernoulli trial with success probability p and failure probability q = 1-p.
- Consider n independent trials, where independence ensures probabilities multiply.
- A specific sequence of k successes and n-k failures has probability p^k q^(n-k).
- Since order does not matter for the total count, multiply by the number of such sequences.
- The number of sequences is given by the binomial coefficient n choose k.
- Summing these probabilities yields the PMF for the total count X.
Worked example
Calculate the probability that exactly 2 out of 5 cells are neutrophils, given p=0.6.
- Identify n=5, k=2, p=0.6, and q=0.4.
- Calculate the binomial coefficient: 5! / (2! * 3!) = 10.
- Calculate the probability of one specific sequence: 0.6^2 * 0.4^3 = 0.02304.
- Multiply the coefficient by the sequence probability: 10 * 0.02304 = 0.2304.
Answer 0.2304
Common mistakes
- Interpreting 'success' as a desirable outcome rather than the event of interest.
- Assuming independence when sampling without replacement from a small population.
- Confusing the probability of exactly k successes with cumulative probabilities like P(X >= k).
- Forgetting to use the complement rule for tail probabilities.