A particular discrete distribution. An experiment or trial with two possible results, usually classified as ‘success’ or ‘failure’, is repeated independently until the first success is obtained. If the probability of a success, p (≠ 0, 1), is the same for each trial, then the distribution of the total number of trials, up to and including the one in which the first success is obtained, is a geometric distribution with parameter p. The probability function is given by
The reason for the term ‘geometric’ is that the sequence of probabilities P(X = r), for r = 0, 1, 2,…, forms a geometric progression with first term p and common ratio (1−p). The mean is 1/p and the variance is (1−p)/p2. The mode is 1 for all values of p.
The fact that the mode is always 1 means that, even for a very rare event (e.g. a lottery win), the most probable number of trials necessary to obtain the first success is 1. This is sometimes regarded as a paradox.
If we write q = 1−p, cumulative probabilities are given by P(X≥r) = qr−1 and P(X≤r) = 1−qr. The geometric distribution is a discrete analogue of the exponential distribution and shares the ‘non-ageing’ or forgetfulness property:
If X1, X2,…, Xn are independent random variables each having a geometric distribution with parameter p, then , which is the number of trials up to and including the nth success, has a negative binomial distribution with parameters n and p.
Some authors take the definition of a geometric distribution to be the number of trials before the first successful trial. This leads to P(X=r)=p(1−p)r, for r=0, 1,…, for which the mean is (1−p)/p and the mode is 0, though the variance is unchanged.