The property $\sqrt[n]{ab} = \sqrt[n]{a}\sqrt[n]{b} $ is valid just if $a,b>0$. If this restriction didn't apply $\sqrt{-1}\sqrt{-1} = -1$ would be equal to $\sqrt{(-1)(-1)} = 1$
Beside giving particular examples, how can I show that the $a,b>0$ is required for $\sqrt[n]{ab} = \sqrt[n]{a}\sqrt[n]{b} $ to be true?
The answer to this question really comes down to the definition of $\sqrt[n]{a}$. If this symbol describes any number $x$ so that $x^n = a$, then $\sqrt[n]a$ describes a set. $\sqrt{-1}\sqrt{-1} = \pm i^2 = \pm 1 = \{-1.1\}$, while $\sqrt{(-1)(-1)} = \sqrt{1} = \pm 1 = \{-1,1\}$. So, in fact, the set of values defined by these two expressions are equivalent.
You may want to define $\sqrt[n]a$ as the unique positive real number $x, x>0$ when $a>0$ and the unique positive imaginary number $ix, x>0$ when $a<0$. Then, if $a,b<0$, then $ab>0$, so $\sqrt[n]{ab}>0$ by definition. $\sqrt[n]a\sqrt[n]b = xi\cdot yi = -xy <0$ on the other hand, hence the discrepancy.
When you put a restriction on the definition for $\sqrt[n]a$, you're going to have restrictions on it's properties as well.