Law of convolution of two distributions

46 Views Asked by At

I'm currently trying to understand the proof that probability distributions are uniquely defined by their characteristic functions.

One of the relevant lemmas starts like this

\begin{aligned} &\text { Proposition } 10.12 \text { Let } P \text { be a probability distribution on } \mathbb{R}^{d} \text { . Let } P^{\sigma}=P*\mathcal{N}\left(0, \sigma^{2} I\right)=\\ &\mathcal{L}(X+\sigma Z) \text { where } X, Z \text { are independent with } \mathcal{L}(X)=P \text { and } \mathcal{L}(Z)=\mathcal{N}\left(0, \sigma^{2} I\right), \sigma>0 \end{aligned}

For two independent random variables the convolution is given by

$$ P(X+Z\in B) = P_X*P_Z(B)$$

I don't understand why, given that $\mathcal{L}(Z)=\mathcal{N}\left(0, \sigma^{2} I\right)$ and $\mathcal{L}(X)=P$, the law for $P^\sigma$ is given by $\mathcal{L}(X+\sigma Z)$ and not $\mathcal{L}(X+Z)$.