Suppose I want to prove that a sequence $(s_n)$ converges, and I don't know much more about $(s_n)$ than a few properties. (That is, I don't know a closed-form formula.)
I have seen a proof written by my professor that began with "we're concerned with only the convergence of $(s_n)$, so suppose $s_0 = 0$."
I believe he was implicitly saying, "without loss of generality, suppose $s_0 = 0$." Certainly there are plenty of sequences with the given properties that don't start at $0$, so it seemed to me that this could not be done. Upon considering it further, though, I believe I may have a theory on the overall argument.
If $(a_n)$ and $(b_n)$ are equivalent, then for any $\epsilon > 0$, there exists some $N$ so that for all $n \geq N$, $|a_n - b_n| \leq \epsilon$. If $(a_n)$ is Cauchy, then $(b_n)$ is Cauchy and vice-versa. Hence, if $(a_n)$ converges, then $(b_n)$ converges, and vice-versa. If the first term is $0$, that will neither impact convergence nor equivalence because we can throw out a finite number of terms anyway. Hence, if I prove a result for an equivalent sequence $(a_n)$ that starts at $a_0 = 0$, the result likewise holds for $(b_n)$.
Because of these, we can in effect say, "without loss of generality, suppose $s_0 = 0$" because proving that an equivalent sequence $(s_n)$ that does start at $0$ convergences also implies that $(s_n)$ converges.
Is this correct? More generally, what are the limits to assuming something ``without loss of generality?''