Random Variables & Their Distributions

the indicator random variable

An indicator random variable is the simplest possible random variable: it just reports whether an event happened. It equals 1 if the event occurs and 0 if it does not, turning a yes/no question into a number. For an event A we write the indicator as 1_A: it is 1 on every outcome inside A and 0 on every outcome outside A. A coin landing heads, a customer making a purchase, a die showing a six, each becomes a crisp 0 or 1.

Because it takes only the values 0 and 1, an indicator has a very tidy distribution: P(1_A = 1) = P(A) and P(1_A = 0) = 1 - P(A). This is exactly a Bernoulli variable with success probability p = P(A). The single most useful fact about it is that its expectation equals the probability of the event: E[1_A] = 1 times P(A) plus 0 times (1 - P(A)) = P(A). The average value of the on-off switch is just how often it is on.

This little variable is a workhorse far out of proportion to its simplicity. To count how many of several events occur, add up their indicators: the total count is a sum of 0-1 variables, and by linearity of expectation its expected value is simply the sum of the individual probabilities, no independence required. That trick (write a count as a sum of indicators) cracks problems that look hard head-on, such as the expected number of matches, fixed points, or empty boxes. Indicators are the bridge that lets you compute with events using the algebra of numbers.

Roll a die three times; let A_i be the event that roll i is a six, with indicator 1_{A_i}. The number of sixes is X = 1_{A_1} + 1_{A_2} + 1_{A_3}. So E[X] = E[1_{A_1}] + E[1_{A_2}] + E[1_{A_3}] = 1/6 + 1/6 + 1/6 = 1/2, instantly.

An indicator turns an event into a 0-1 number whose expectation is the event's probability.

The defining identity is E[1_A] = P(A): the expectation of an indicator is the probability of its event. Combined with linearity of expectation, this lets you find expected counts even when the events are dependent.

Also called
indicatorindicator functionBernoulli indicator0-1 variable指示變數示性函數指標變數