Let $p$ a prime number, ${q_{_1}}$,..., ${q_{_r}}$ are the distinct primes dividing $p-1$, ${\mu}$ is the Möbius function, ${\varphi}$ is Euler's phi function, ${\chi}$ is Dirichlet character $\bmod{p}$ and ${o(\chi)}$ is the order of ${\chi}$.
How can I show that: $$\sum\limits_{d|p - 1} {\frac{{\mu (d)}}{{\varphi (d)}}} \sum\limits_{o(\chi ) = d} {\chi (n)} = \prod\limits_{j = 1}^r {(1 - \frac{1}{{\varphi ({q_j})}}} \sum\limits_{o(\chi ) = {q_{_j}}} {\chi (n)} ) \quad ?$$
Fix $n$ and define
$$f(d)=\sum_{o(\chi)=d}\chi(n).$$
Let's show $f$ is multiplicative. First off, let $g$ be a generator for $(\mathbb{Z}/p\mathbb{Z})^\times$ and write $n=g^k$, then let $\psi$ be a generator for the group $\{\chi:\chi^d=1\}$, in which case we may say $o(\chi)=d\iff \chi=\psi^e$ for a unit $e$ mod $d$. As $\psi(g)$ is a primitive $d$th root of unity, the values $\psi(g)^e$ (as $e$ ranges over units mod $d$) will be all primitive $d$th roots of unity, i.e. all $\zeta$ with $o(\zeta)=d$. Thus
$$ f(d)=\sum_{(m,d)=1} \psi^m(g^k)=\sum_{o(\zeta)=d}\zeta^k. $$
This is known as a Ramanujan sum. We want to see that $f(d_1d_2)=f(d_1)f(d_2)$ whenever $d_1,d_2$ are coprime. (Technically since $k$ is an integer mod $p-1$ the formula only makes sense for when $d\mid(p-1)$, however if we interpret it as a function of a usual integer $k$ we can talk about any $d$ values we want and it will in fact be multiplicative).
The key is the Chinese Remainder Theorem $(\mathbb{Z}/d_1d_2\mathbb{Z})^\times\cong(\mathbb{Z}/d_1\mathbb{Z})^\times\times(\mathbb{Z}/d_2\mathbb{Z})^\times$ which for our purposes means every $d_1d_2$th root of unity is uniquely expressible as a product of $d_1$th and $d_2$th roots of unity, and in particular primitive $d_1d_2$th roots of unity are uniquely expressible as products of primitive $d_1$th and $d_2$th roots of unity. Therefore,
$$ \begin{array}{ll} f(d_1)f(d_2) & \displaystyle =\left(\sum_{o(\zeta)=d_1}\zeta^k\right)\left(\sum_{o(\xi)=d_2}\xi^k\right) \\[7pt] & \displaystyle =\sum_{\substack{o(\zeta)=d_1 \\ o(\xi)=d_2}} (\zeta\xi)^k =\sum_{o(\eta)=d_1d_2}\eta^k \\[7pt] & =f(d_1d_2). \end{array} $$
Now, if $g(m)$ is any multiplicative function then we may write $g(m)=\prod_{q^r\| m}g(q^r)$, where $q^r\|m$ means $q^r$ is the power of a prime $q$ that appears in $m$'s prime factorization. Moreoever, if $g(m)$ is multiplicative then so is $\sum_{d\mid m}g(d)$ as a function of $m$, in which case it also gets a factorization that looks like $\prod_{q^r\|m} \sum_{d\mid q^r}g(d)$. If $g(d)$ has a $\mu(d)$ factor, then $\sum_{d\mid q^r}g(d)=g(1)+g(q)$ because $g(q^r)=0$ if $r>1$. Therefore we have
$$ \sum_{d\mid(p-1)}\frac{\mu(d)}{\varphi(d)}f(d)=\prod_{q\mid (p-1)}\left(1-\frac{1}{\varphi(q)}f(q)\right) $$
as desired.