1. Types of Convergence for Random Variables
When dealing with sequences of random variables, understanding how they behave as the number of variables increases is crucial. This concept is known as convergence. There are several distinct modes of convergence, each with specific implications.
1.1 Convergence in Probability
A sequence of random variables {Xn} converges in probability to a random variable X (denoted as Xn →P X) if for every ε > 0,
limn→∞ P(|Xn - X| ≥ ε) = 0
Explanation: This means that as n gets very large, the probability that Xn is significantly different from X becomes arbitrarily small. It implies that the values of Xn are likely to be close to X, but there might still be occasional large deviations.
Example: Consider a sequence of indicator random variables where Xn = 1 if a fair coin lands heads (probability 0.5) and Xn = 0 if it lands tails, for n trials. If we define X as a constant 0.5, then the sample mean of these indicators, which estimates the probability of heads, converges in probability to 0.5.
1.2 Almost Sure Convergence
A sequence of random variables {Xn} converges almost surely to a random variable X (denoted as Xn →a.s. X) if
P(limn→∞ Xn = X) = 1
Explanation: This is a stronger form of convergence. It means that the set of outcomes for which Xn does not converge to X has probability zero. In simpler terms, for almost all possible outcomes in the sample space, the sequence of random variables Xn will eventually settle down and equal X. It's like saying the sequence converges