The Jacobian: Mapping Transformations

The Intuition: The Stretched Rubber Band

Imagine a rubber band with a uniform distribution of dots. If you stretch the rubber band, the dots move further apart, and the density of the dots at any one point decreases.

In calculus, we use the derivative to measure how much a function is changing. In probability, we use the absolute value of the derivative—the Jacobian—to "re-normalize" the PDF after it has been stretched.

1D Transformation Rule

If $Y = g(X)$ is a strictly monotonic (always increasing or always decreasing) and differentiable function, and $X$ has the PDF $f_X(x)$, the PDF of $Y$ is:

Breaking it Down:

1. $f_X(g^{-1}(y))$: This part tells us: "What was the probability of $X$ at the point that mapped to this $y$?"
2. $\left| \frac{d}{dy} g^{-1}(y) \right|$: This is the Jacobian Factor. It is the "stretching factor." It compensates for how much the transformation changed the "width" of our probability intervals.

Why is it Absolute Value?

Probability densities ($f(x)$) can never be negative. Since a function $g(X)$ might be decreasing, its derivative would be negative. The absolute value ensures that our new density remains a positive, valid PDF.

Multi-Dimensional (The Jacobian Determinant)

When we move to Joint Distributions (e.g., $(X, Y) \to (U, V)$), we don't just use a single derivative. We use a Jacobian Matrix of partial derivatives, and we take its Determinant.

The determinant measures how much the 2D area (or 3D volume) has been scaled by the transformation. This is the foundation of change-of-variables in multi-variable calculus and high-dimensional statistics.

Test Your Knowledge