5
$\begingroup$

Let $X \sim \mathcal{N}(\mu, \sigma^2)$. Are there any (1) general formula and (2) references to the general formula for

$$ \mathbb{E} (X^n e^{tX}),\; n \in \mathbb{N}, t \in \mathbb{R}$$

in particular for $n = 1$ and $n = 2$?


I know $\mathbb{E}(X^n)$ are the normal moments ($\mu$, $(\sigma^2 + \mu^2)$, ... for increasing $n$). I also know $\mathbb{E}(e^{tX})$ is the moment generating function for a normal, which evaluates to

$$ \mathbb{E}(e^{tX}) = \exp\left(\mu t + \frac{1}{2}\sigma^2 t^2\right).$$

Clearly, directly multiplying both will not work as the two parts are dependent. I also considered Delta method and making $X^n e^{tX}$ a derivative of something and using the exchangeability of expectation and derivatives, but few pages of calculations in and it does not look promising.

I also looked at the table of normal integrals by Owen (1980), but am unable to find anything of the form (ignoring the constants)

$$ \int x^n \exp(tx) \exp\left(-\frac{1}{2}\frac{(x-\mu)^2}{\sigma^2}\right) \textrm{d}x .$$

$\endgroup$
2
  • 1
    $\begingroup$ Just complete the square to reduce this to a question about the moments of a Normal distribution. $\endgroup$
    – whuber
    Commented Oct 21, 2020 at 13:14
  • 1
    $\begingroup$ I have to admit I have not considered that - in an (admittedly more complex) parameterisation I was too determined to take things out the last exponential function without considering whether I should put things in. Thanks for the tip, will report back. $\endgroup$
    – B.Liu
    Commented Oct 21, 2020 at 14:25

2 Answers 2

2
$\begingroup$

As hinted by @whuber (with thanks), the key is to transform the expectation to the (pure) moment of another normal distribution.

We first recognise the expectation in question is, by definition

$$ \mathbb{E}_{X \sim \mathcal{N}(\mu, \sigma^2)}(X^n e^{tX}) = \int x^n \exp(tx)\frac{1}{\sqrt{2\pi \sigma^2}}\exp\left(-\frac{1}{2}\frac{(x-\mu)^2}{\sigma^2}\right) \,\textrm{d}x .$$

We combine the exponential terms on the RHS to obtain

$$ \int x^n \frac{1}{\sqrt{2\pi \sigma^2}} \exp\left(-\frac{1}{2}\left[\frac{(x-\mu)^2}{\sigma^2} - 2tx\right]\right) \,\textrm{d}x $$

Completing the square within the square brackets we have

$$ \int x^n \frac{1}{\sqrt{2\pi \sigma^2}} \exp\left(-\frac{1}{2}\left[\frac{(x-(\mu+t\sigma^2))^2}{\sigma^2}\right]\right) \exp\left(\mu t + \frac{1}{2}\sigma^2 t^2\right) \,\textrm{d}x $$

The rightmost exponential term can be moved out of the integral. What is left inside is the moment of a different normal $\mathcal{N}(\mu + t\sigma^2, \sigma^2)$ by definition. Thus

$$ \mathbb{E}_{X \sim \mathcal{N}(\mu, \sigma^2)}(X^n e^{tX}) = \mathbb{E}_{X \sim \mathcal{N}(\mu + t\sigma^2, \sigma^2)}(X^n) \cdot \exp\left(\mu t + \frac{1}{2}\sigma^2 t^2\right) $$

(For completeness) for $n = 1, 2$,

$$\mathbb{E}_{X \sim \mathcal{N}(\mu + t\sigma^2, \sigma^2)}(X) = \mu + t\sigma^2$$ $$\mathbb{E}_{X \sim \mathcal{N}(\mu + t\sigma^2, \sigma^2)}(X^2) = Var(X) + \mathbb{E}^2_{X \sim \mathcal{N}(\mu + t\sigma^2, \sigma^2)}(X) = \sigma^2 + (\mu+t\sigma^2)^2.$$

$\endgroup$
2
$\begingroup$

Let $X$ be a random variable with moment generating function (mgf) $m(t),$ where $t$ is a real number. Then, $$E[X^n exp(tX)] = E[\frac{d^n}{dt^n} exp(tX)] = \frac{d^n}{dt^n} m(t),$$ which is the n-th derivative with respect to $t$ of $m(t)$ and $E$ denotes expectation.

In this specific example, $m(t)$ is the mgf of a normal random variable.

$\endgroup$
1
  • $\begingroup$ Please use $\LaTeX$ on this site, as it makes reading mathematics much easier. $\endgroup$ Commented Jun 7, 2022 at 14:20

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.