Prove: $\int_{1}^{t} (t-s)^{-\alpha} s^{-\beta}ds \leq Ct^{-\alpha}$, where $0<\alpha<1<\beta$ and $C$ is a constant, $t \geq 1$.

Progress: I already tried two ways to show it. One is divide $t^{-\alpha}$ on both sides and tried to approach Beta distribution however I failed since one of the parameters would be smaller than zero, rather than bigger than zero. Another way is after dividing like in the first method, I tried to bound the integral directly but it seems still not be able to be bounded. I appreciate any comments or solutions.


Observe \begin{align} \int^t_1 (t-s)^{-\alpha}s^{-\beta}\ ds =&\ \int^t_{t/2+1/2} (t-s)^{-\alpha} s^{-\beta}\ ds + \int^{t/2+1/2}_{1} (t-s)^{-\alpha} s^{-\beta}\ ds\\ =:&\ I_1+I_2. \end{align} For $I_1$, we have \begin{align} I_1 \leq \frac{1}{(t/2+1/2)^\beta} \int^t_{t/2+1/2} \frac{ds}{(t-s)^{\alpha}} \leq C \frac{(t-1)^{1-\alpha}}{(t+1)^\beta} \leq C't^{1-\alpha-\beta} \end{align} when $t\gg 1$.

For $I_2$, we have \begin{align} I_2 \leq C\frac{1}{(t/2-1/2)^\alpha} \int^{t/2+1/2}_1 s^{-\beta}\ ds \leq C't^{-\alpha} \end{align} when $t\gg 1$.

Hence it follows \begin{align} I_1 + I_2 \leq C_1t^{1-\alpha-\beta}+C_2t^{-\alpha} \leq C_3t^{-\alpha}. \end{align}

  • $\begingroup$ Could you be more specific about estimating $I_1$ and$I_2$, especially $C'$? $\endgroup$ – Topoguy Oct 25 '16 at 21:47
  • $\begingroup$ @JackD'Aurizio They don't depend on $t$. Observe for $I_1$, it's not hard to show that $C=1-\alpha$ since $(t-1)^{1-\alpha} \leq (t+1)^\beta$ for $t\gg 1$. And for $I_2$, we can see that $C = \int^\infty_1 \frac{ds}{s^\beta}$. $\endgroup$ – Jacky Chong Oct 25 '16 at 22:42
  • $\begingroup$ @JackyChong: I owe you some apologies, since in my observation I set $\beta=1$, that is not really allowed by the given constraints. Splitting in halves is perfectly fine. (+1) and sorry if I caused you an unnecessary time loss. $\endgroup$ – Jack D'Aurizio Oct 25 '16 at 23:01
  • $\begingroup$ @JackD'Aurizio No apology necessary. We are only trying to get things right. $\endgroup$ – Jacky Chong Oct 25 '16 at 23:05
  • $\begingroup$ @JackD'Aurizio Hi, where is $t^{1-\beta}$ in your answer? I mean you are trying to show $t^{1-\beta}\int_{1/t}^{1}(1-u)^{-\alpha}u^{-\beta}\,du\color{red}{\leq} C$, but I only see you just showed the bound for the integral. $\endgroup$ – Topoguy Oct 26 '16 at 0:21

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.