Lagmental Vicfred

Bayes' Formula Reverses a Conditional Direction by Vicfred

A posterior probability combines a likelihood with its prior and normalizes over all hypotheses. I want the notation, the mechanism, and the failure mode visible at the same time.

Set-up

A probability space \((\Omega,\mathcal F,\mathbf P)\) separates outcomes, measurable events, and their probabilities. A random variable \(X:\Omega\to\mathbf R\) must be measurable.

$$ \{H_i\}_{i=1}^{m}\ \text{a partition},\qquad\mathbf P(E)>0 $$

The first display fixes the mathematical data. I label it \(\mathsf{data}\) mentally, while the next is the \(\mathsf{claim}\); the bridge between them is the displayed \(\Longrightarrow\), not an automatic implication.

$$ \mathbf P(H_j\mid E)=\frac{\mathbf P(E\mid H_j)\mathbf P(H_j)}{\sum_i\mathbf P(E\mid H_i)\mathbf P(H_i)} $$

The calculation

Here is a concrete symbolic test. Reading it from left to right reveals which transformation is reversible and which is only an implication.

$$ \underbrace{\mathbf P(H_j\mid E)}_{\text{posterior}}\propto\underbrace{\mathbf P(E\mid H_j)}_{\text{likelihood}}\underbrace{\mathbf P(H_j)}_{\text{prior}} $$

What survives abstraction

The aligned summary deliberately puts the datum and conclusion on different rows. Mathematically, this is the distinction between specifying an object and proving a property of it.

$$ \begin{aligned} \mathsf{D}\;&:\quad \{H_i\}_{i=1}^{m}\ \text{a partition},\qquad\mathbf P(E)>0,\\[5pt] \mathsf{C}\;&:\quad \mathbf P(H_j\mid E)=\frac{\mathbf P(E\mid H_j)\mathbf P(H_j)}{\sum_i\mathbf P(E\mid H_i)\mathbf P(H_i)}. \end{aligned} $$

The boundary

Conditioning on a probability-zero event cannot be done by naïvely dividing. Conditional densities and regular conditional probabilities require additional structure.

$$ \boxed{\begin{gathered} \text{compact conclusion}\\[-2pt] \mathbf P(H_j\mid E)=\frac{\mathbf P(E\mid H_j)\mathbf P(H_j)}{\sum_i\mathbf P(E\mid H_i)\mathbf P(H_i)} \end{gathered}} $$

I would use the boxed line as a reference later, while returning to the full display whenever a hypothesis becomes uncertain. That division keeps compression from becoming ambiguity.

This article was posted on Sun 17 May 2015. Facts and circumstances may have changed since publication.
Please contact me before jumping to conclusions if something seems wrong or unclear.