Let $X_1$ and $X_2$ be independent Normal random variables with mean $\mu_1$ and $\mu_2$, and variances $\sigma_1$ and $\sigma_2$. Let $Y = X_2-X_1 + c$, where $c$ is a constant.

For notational simplicity, let $Y = X' + c$, where $X' = X_2-X_1$

The moment generating function of a Normal random variable $X_n\sim N(μ_n,σ_n^2)$ is $$M_{X_n } (t)=\text{exp}\Big(μ_n t+\frac{σ_n^2 t^2}{2}\Big)$$

Then, $$M_{X'} (t)=\text{exp}\Big[t(\mu_1+\mu_2)+\frac{t^2}{2}(\sigma^2_1+\sigma^2_2)\Big]$$

Thus, by the uniqueness property of moment generating function, $X'$ is said to follow a Normal distribution with mean $(\mu_1+\mu_2)$ and variance $(\sigma^2_1+\sigma^2_2)$.

How do I proceed from here to derive the following:

  1. First, I want to establish that $Y$ is also a Normal random variable using the moment generating function. The constant $c$ bugs me, but I cannot ignore it. As per my understanding, $Y\sim N\big(c+\mu_1+\mu_2,\sigma^2_1+\sigma^2_2\big)$ (and I could be completely wrong). But I don't understand how to apply the moment generating function to prove that.

  2. Secondly, I want to derive $P(Y \leq a)$, where $a$ is a constant. If I am not mistaken, $P(Y \leq a)$ can be written as $$P(Y \leq a)=\Phi\Bigg(\frac{a-X_2-X_1 + c}{\sqrt{2(\sigma^2_1+\sigma^2_2)}}\Bigg).$$ But how do I arrive at that.

Any help is high solicited and appreciated.


It should be $\mu_2-\mu_1$ everywhere, not $\mu_2+\mu_1$. For the constant, you have $M_c(t)=e^{ct}$ and when multiplied with $M_{X'}(t)$, you have $$M_Y(t)=\exp(t(c-\mu_1+\mu_2)+...)$$ So, by uniqueness property, you show $Y$ is normal.

For the second, $$P(Y\leq a)=P\left(\underbrace{\frac{Y- \mu_Y}{\sigma_Y}}_{\text{ Standard normal}}\leq \frac{a-\mu_Y}{\sigma_Y}\right)=P\left(Z\leq\frac{a-\mu_Y}{\sigma_Y}\right)=\Phi\left(\frac{a-\mu_Y}{\sigma_Y}\right)$$

where $\sigma_Y=\sqrt{\sigma_1^2+\sigma_2^2}, \mu_Y=c+\mu_2-\mu_1$.

  • 1
    $\begingroup$ Thank you very much @gunes for the prompt response and for pointing out the error in the mean. $\endgroup$
    – nashynash
    Sep 7 '20 at 13:15

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.