9
$\begingroup$

Suppose, $X\sim N(\mu,\sigma^2)$.

Can anyone help in finding the following : $\text{Var } \bigg(\Phi\big(\frac{X + c}{d}\big) \bigg)$ ? Here, c and d are positive.

Here, $\Phi(x)$ is the "Cumulative Distribution Function" of the above-mentioned normal distribution.

Thanking you in advance.

$\endgroup$
1
  • 5
    $\begingroup$ Look at this answer for a technique that will likely help you find $E\left[\Phi\left(\frac{X+c}{d}\right)\right]$. Perhaps the same idea can be adapted to find $E\left[\Phi^2\left(\frac{X+c}{d}\right)\right]$, from which two results you can get $\text{Var}\left(\Phi\left(\frac{X+c}{d}\right)\right)$. $\endgroup$ Commented May 13, 2016 at 19:22

1 Answer 1

4
$\begingroup$

Expanding on the comment by Dilip Sarwate, using this list of integrals of Gaussian functions gives

\begin{align*} E \left[ \Phi\left(\frac{X+c}{d}\right)\right] & = \int \frac{1}{\sigma} \phi\left(\frac{x-\mu}{\sigma}\right) \, \Phi\left(\frac{x+c}{d}\right) \, dx \\[8pt] & = \int \phi\left(x\right) \, \Phi\left(\frac{\sigma x+ \mu + c}{d}\right) \, dx\\[8pt] & = \Phi \left( \frac{\mu+c}{\sqrt{\sigma^2 + d^2}}\right) \end{align*} and

\begin{align*} E \left[ \Phi\left(\frac{X+c}{d}\right)^2\right] & = \int \frac{1}{\sigma} \phi\left(\frac{x-\mu}{\sigma}\right) \, \Phi\left(\frac{x+c}{d}\right)^2 \, dx \\[8pt] & = \int \phi\left(x\right) \, \Phi\left(\frac{\sigma x+ \mu + c}{d}\right)^2 \, dx\\[8pt] & = \Phi \left( \frac{\mu+c}{\sqrt{\sigma^2 + d^2}}\right) - 2T \left( \frac{\mu+c}{\sqrt{\sigma^2 + d^2}}, \frac{d}{\sqrt{2\sigma^2+d^2}}\right) \end{align*} where $T$ is Owen's T function.

From these expressions, $\text{Var}\left( \Phi\left(\frac{X+c}{d}\right) \right)$ follows.

$\endgroup$
1
  • 1
    $\begingroup$ Looks healthy ! $\endgroup$
    – wolfies
    Commented Sep 28, 2016 at 16:57

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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