2
$\begingroup$

The variable $X$ has a continuous probability density function (for example, it may be normally distributed) with mean $a+b$ and a constant variance, say $1$.

I try to find the following:

$P(ab \geq 0 | X=x)$

Using the Bayes' theorem, can I turn it into this:

$\frac {p(X=x | ab \geq 0) P(ab \geq 0)} {p(X=x)}$?

I find that for it to make sense, I could not have had $P(X=x | ab \geq 0)$ and $P(X=x)$ because otherwise both the numerator and the denominator would be 0, so it has to be $p(X=x | ab \geq 0)$.

I also compared the units of the LHS and the RHS. Probability is a pure number, and let $X$ have the unit, "$unit$", then probability density has the unit "$\frac {1} {unit} = unit^{-1}$".

For the LHS,

$P(ab \geq 0 |X=x) = \frac {P(ab \geq 0, X=x)} {p(X=x)}$. The numerator would have the unit $1•unit^{-1} = unit^{-1}$ and the denominator would have the unit "$unit^{-1}$". These cancel out to give us a pure number as expected.

For the RHS,

The numerator again has the unit $unit^{-1}$ and the denominator has the unit $unit^{-1}$, cancelling out again to give us a pure number.

This hints that the expression is correct.

Is it?

$\endgroup$

1 Answer 1

1
$\begingroup$

Yes, the expression (and your intuition) is correct. To put it more clearly, let $A$ be an arbitrary event: $$P(A|X=x)=\frac{p_X(x|A)P(A)}{p_X(x)}$$

It's generally useful to think about the approximation case where $P(X=x)\approx p_X(x)dx$, where you can find the above expression algebraically.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.