Probability

A probability density is simply a curve that shows where an uncertain quantity, like a waiting time, is likely to land: tall where outcomes are common, low where they're rare. ("Density" because it tells you how packed together the chance is near each value, the way population density tells you how packed together people are.) All we're really saying is that chance can be measured as area under that curve, which is how you'd answer "what are the odds I wait more than 10 minutes?" $P(a \le X \le b) = \int_a^b f(x)\,dx$ looks formal, but it's just $X$ (the thing you're unsure about, like how long you'll wait), $f(x)$ (the density curve itself), and the integral from $a$ to $b$ (the area under that curve between two values, which is the chance of landing there). A few simple parts; bear with the notation.

Think of a histogram of, say, how long 1000 callers waited on hold. Each bar's area is the fraction of callers in that range, so the areas add to 1. Now imagine millions of callers and thinner and thinner bars. The tops of the bars smooth out into a curve $f(x)$, and "fraction of callers between $a$ and $b$" turns into the area under that curve from $a$ to $b$. That's an integral.

So a probability density function satisfies $f(x) \ge 0$ and $\int_{-\infty}^{\infty} f(x)\,dx = 1$, and the probability that $X$ lands in an interval is the area over that interval.

A common mistake. It's tempting to read $f(3)$ as "the probability that $X = 3$." It isn't. $f$ is a density, probability per unit of $x$, and it can even be bigger than 1. Probability only comes from area, and a single point has no width, so $P(X = 3) = \int_3^3 f = 0$. Only intervals get positive probability.

The mean. The average value of $X$ is $\mu = \int x f(x)\,dx$. Look familiar? It's the same weighted average as $\bar x$ in the center of mass lesson, with $f(x)\,dx$ playing the role of each strip's share of the area. So the mean is the point where the density curve would balance.

The exponential distribution models waiting times (the next customer, the next decay of an atom):

The normal distribution is the bell curve,

It has no elementary antiderivative, so its probabilities come from tables or numerical integration, just like the arc lengths that wouldn't simplify.

The median $m$ splits the area in half: $\int_{-\infty}^{m} f(x)\,dx = \tfrac12$. For a lopsided density like the exponential, the median and the mean differ, because a few very long waits drag the balance point out to the right.

Reliability engineers model part lifetimes with exponential densities, insurers price policies from loss distributions, and every polling margin of error comes from areas under the normal curve. The Galton board lab shows the bell curve emerging from coin flips.