Let $f(n) = n^2 , g(n) = n$, and $h(n) = g(n)−f(n)$. It is clear that $h(n) ≤ g(n) ≤ f(n)$ for all $n ≥ 0$. Therefore, $f (n) = \max(f (n), h(n))$. Thus, $O(n) = O(g(n)) = O(f(n) + h(n)) = O(\max(f(n), h(n))) = O(f(n)) = O(n^2)$.
Can you explain why this statement is wrong?
You're basically claiming that $O(n^2+n-n^2)$ is the same thing as $O(n^2)$, and that's not true.
Depends on what you mean when you write $O(a(n)) = O(b(n))$.
Do you mean that the two set of functions are identical or are you using the (common) abuse of notation to mean $O(a(n)) \subseteq O(b(n))$?
In the former case this is wrong: $O(f(n) + h(n)) = O(\max(f(n), h(n)))$. Notice that the first term is $O(n)$ by definition of $h(n)$, while the second term is $O(n^2)$ since the maximum is always attained by $f(n)$. Clearly, the sets of functions $O(n)$ and $O(n^2)$ do not coincide. For example $n^{3/2} \not\in O(f(n) + h(n))$ but $n^{3/2} \in O(\max(f(n), h(n)))$.
In the second case nothing is wrong. In fact, it is true that $O(n) \subseteq O(n^2)$.
Let $d(n) \in O(n)$ and notice that, by definition of $O(\cdot)$, there exist two constants $n_0 \ge 1$ and $c > 0 $ such that $\forall n \ge n_0$, $d(n) \le c n$. But then $d(n) \le c n \le c n^2$, showing that $d(n) \in O(n^2)$.
Because $h(n)$ is negative, as soon as $n\geq 2$ your argument does not work. Complexities are supposed to be nonnegative and $a(n)=O(b(n))$ requires $a(n)\leq cb(n)$ where $c>0$ is a constant.
