Let $$E = \{(x_1,\ldots,x_n) \mid x_1,\dots,x_n>0, x_1+\cdots+x_n<1 \}$$ and $f:[0,1]\to \mathbb{R}$ be continuously differentiable. Prove $$\int \cdots \int_E f(x_1+ \cdots +x_n) \, dx_1\cdots dx_n=\frac{1}{(n-1)!}\int_0^1 f(s)s^{n-1} \, ds$$
I think induction is the way to go here, the base was rather trivial but now I'm stuck. Any help?
Let $E_n$ be the $n$-simplex and consider $g(s):=\int_0^{1-s} f(s+t) \, dt$. Assuming the claim is true for $n$, we have
\begin{align} & \int_{E_{n+1}} f(x_1+\dots+x_n+x_{n+1}) \, dx_1\dots dx_n \, dx_{n+1}\\ = {} & \int_{E_n}\int_0^{1-({x_1+\dots+x_n})} f(x_1+\dots+x_n+x_{n+1})\,dx_{n+1} \, dx_1\dots dx_n \\ = {} & \int_{E_n} g(x_1+\dots+x_n)\,dx_1\dots dx_n\\ = {} & \frac{1}{(n-1)!} \int_0^1 g(s)s^{n-1}\,ds \end{align} and the rest just follows from integration by parts: \begin{align} \int_0^1 g(s) s^{n-1} \, ds= \frac{1}{n}g(s)s^n \bigg|_0^1 - \frac{1}{n} \int_0^1 g'(s)s^n ds \end{align} The first term is zero because $g(1)=0$. The second term is $$\frac{1}{n}\int_0^1 f(s)s^n \, ds$$ as desired.