LessonThis is a lesson restated in new words. Any figure in it is an example from the source, not this desk's record.
A random variable is a real-valued function on the outcomes of a random experiment. It does not predict the outcome. It gives each outcome a number so you can talk about probabilities, expectations, and later about processes indexed by time. For two coin tosses scored as the number of heads, zero and two each have probability 1/4 and one head has probability 1/2, even though the four raw sequences are equally likely. The map, not the sample space alone, creates those weights.
Why it matters
Last week’s lecture, in this course, rebuilt the probability axioms, counting, conditional probability, partitions, and Bayes’ theorem. This week separates two ideas students glue together. Independent events multiply: the probability both happen equals the product of the probabilities, so conditioning on one leaves the other unchanged. Mutually exclusive events cannot happen together, so the intersection is empty and its probability is zero. If both events have positive probability, they cannot be both exclusive and independent, because a product of positive numbers is not zero. Pairwise independence is not enough for a whole collection. Mutual independence requires that every subcollection multiplies, not only each pair. Markov chains will need that language, because the statements stack.
The application note is the same as in the probability review. You are not forecasting a face. You are computing an expectation. If the average is on your side often enough, the business can make money. A model that “predicts” in the marketing sense has usually skipped this definition.
The idea in plain words
Write two tosses as HH, TH, HT, TT. Score heads as 1 and let X be the sum. Then X is 2, 1, 1, and 0 on those four sequences. If the sequences are equally likely, P(X equals 0) and P(X equals 2) are each 1/4, and P(X equals 1) is 1/2. Equal likelihood of atoms does not mean every value of X is equally likely. The map merges outcomes.
A discrete random variable takes finitely many values, or a countable list such as the number of tosses until the first head. Its mass function gives P(X equals x), nonnegative weights that sum to 1.
A continuous random variable fills an interval. A waiting time can be 1.01. It does not jump from one minute to two as separate atoms. A density is nonnegative and integrates to 1. The probability of an interval is the area under the density. The height at a single point is not the probability of that point, which is zero. More height only means more mass nearby.
The cumulative distribution function is P(X less than or equal to x). It rises to 1 as x goes to infinity and starts from 0 on the far left. A discrete cumulative is a step function. The lecture’s Bernoulli sketch steps by 0.3 and then by 0.7. A continuous cumulative is the integral of the density up to x, and differentiating it returns the density. Either object identifies the law. Named families, starting with a Bernoulli trial of success probability p, come only after that map is clear.
How to use it
When you meet a new symbol X, write the experiment, the sample space, and the rule that turns an outcome into a number. If the values are a list, write a mass function and check that it sums to 1. If the values fill an interval, write a density and integrate the region you care about. Never read the height of a bell curve at a point as “the chance of that exact number.”
Before a stochastic-process chapter, test independence against mutual exclusion on a pair of events with positive probability. If you cannot say which one you mean, the conditional statements in a chain will be noise.
What usually goes wrong
Calling the density a probability. Assuming equally likely outcomes imply equally likely values of X. Treating pairwise independence as mutual independence. And using “predict” for an expectation. The lecture’s correction is that a correct average still leaves the next outcome unknown.
Questions people ask
Why is one head more likely than two heads in two fair tosses?
Two of the four sequences map to exactly one head, and only one sequence maps to two. The random variable merges outcomes, so the masses are 1/2 and 1/4.
What is the probability that a continuous variable equals 1.7 exactly?
Zero, in this setup. You compute the probability of an interval by integrating the density. The density at 1.7 only says how the mass is concentrated nearby.
Can two possible events be mutually exclusive and independent?
Not if each has positive probability. Exclusivity forces the joint probability to zero, and independence would force it to equal a positive product.