1
$\begingroup$

Assume that $N(t)$ is a stochastic process with positive and integer values. We know that:

$\frac{N(t)}{t} \xrightarrow{\text{P}} c, \hspace{0.5cm} t \rightarrow \infty$

where $c$ is positive and integer valued random variable. I would like to prove that: $\frac{[c t]}{N(t)} \xrightarrow{\text{P}} 1$ in a formal way, because it looks intuitively.

Is this attempt even close to the answer?

\begin{align} P \left( \left| \frac{[c t]}{N(t)} - 1 \right| > \epsilon \right) &= P \left( [c t] - N(t) > \epsilon N(t) \right) + P \left( [c t] - N(t) < - \epsilon N(t) \right) \\&\leq P \left( ct + 1 - N(t) > \epsilon N(t) \right) + P \left( ct - 1 - N(t) < -\epsilon N(t) \right)\\ & = P \left( ct - N(t) > \epsilon N(t) - 1 \right) + P \left( ct - N(t) < \epsilon N(t) + 1 \right). \end{align}

And what can I do now for obtaining the proof?

$\endgroup$
  • $\begingroup$ Note that $x - 1 \leq [x] \leq x$ $\endgroup$ – BGM Aug 10 '18 at 16:10
  • $\begingroup$ Would you recommend try do it with definition of the convergence in probability? $\endgroup$ – FNTE Aug 10 '18 at 16:12
0
$\begingroup$

Maybe it could be simpler to do a reasoning with sequences. It suffices to show that for each sequence $\left(t_n\right)_{n\geqslant 1}$ such that $\lim_{n\to +\infty}t_n=+\infty$, the sequence $\left(Y_n\right)_{n\geqslant 1}$ defined by $$ Y_n:=\frac{\left[ct_n\right]}{N\left(t_n\right)} $$ converges to $1$ in probability. We can use the following facts.

  1. If a sequence of positive random variables $\left(Z_n\right)$ converges to $1$ in probability, so does the sequence $\left(1/Z_n\right)$.
  2. If the sequences of positive random variables $\left(Z_n\right)$ and $\left(Z'_n\right)$ converges to $1$ in probability, so does $\left(Z_nZ'_n\right)$.
  3. $\left(\left[ct_n\right]/\left(ct_n\right)\right)_{n\geqslant 1}$ converges to $1$ in probability.
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.