Characteristic function of a sum of independent variables (source code)

= Characteristic function of a sum of independent variables

Independence makes the characteristic function of a sum equal the product of the individual characteristic functions.

For mutually <independent random variables> $X_1,\ldots,X_n$, write $S_n=\sum_{j=1}^nX_j$. Then
$$
\varphi_{S_n}(t)=\mathbb E\left[\prod_{j=1}^ne^{itX_j}\right]
=\prod_{j=1}^n\mathbb E[e^{itX_j}]
=\prod_{j=1}^n\varphi_{X_j}(t).
$$
This is the <expectation of a product of independent random variables> applied to bounded <complex exponentials>; no <moment> hypothesis is needed.<Probability: Theory and Examples (fifth edition) by Rick Durrett>[§3.3.1, Theorem 3.3.2 and proof, printed p. 126]{ref}

Identical <probability distributions> are unnecessary. When the <random variables> are <independent and identically distributed random variables>, the product is $\varphi_X(t)^n$. A deterministic rescaling also follows directly from the definition: $\varphi_{aX+b}(t)=e^{itb}\varphi_X(at)$ for real $a,b$. Consequently the <sample mean> has <characteristic function> $\varphi_X(t/n)^n$, and a sum divided by $\sqrt n$ has <characteristic function> $\varphi_X(t/\sqrt n)^n$.

The <independence> assumption does real work. Let $X$ take $-1$ and $1$ with equal <probabilities> and set $Y=X$. Both <characteristic functions> are $\cos t$, but $\varphi_{X+Y}(t)=\cos(2t)$, not $\cos^2t$. At $t=\pi/2$ they give $-1$ and zero respectively. Knowing the two marginal <probability distributions> does not determine the sum without their <joint probability distribution>.