Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

If I have $x_1, x_2,\ldots, x_n$ independent NON-identically distributed Bernoulli random variables, how do I show that: $$\mathrm{Pr}\left(\sum_{i=1}^nx_i>\beta\mu\right)\le e^{-g(\beta)\mu}$$

where $$\beta>1$$$$\mu=E\left(\sum_{i=1}^nx_i\right)$$$$g(\beta)=\beta\times \ln(\beta)-\beta+1$$? I believe this can be accomplished using the Markov inequality (because that's what we've been covering), but I'm still not sure how to apply it.

share|cite|improve this question

In markov's inequality, Note that

$$P(\sum_{i=1}^n x_i > \beta \mu) = P(\beta^{\sum_{i=1}^n x_i} > \beta^{\beta \mu})$$

Now apply markov inequality to rhs to get

$$\leq \frac{E[\beta^{\sum_{i=1}^n x_i}]}{\beta^{\beta \mu}} = \frac{\prod_{i=1}^nE[\beta^{x_i}]}{\beta^{\beta \mu}}=\frac{\prod_{i=1}^nE[\beta^{x_i}]}{e^{(\beta \ln \beta) \mu}}$$

Now $E[\beta^{x_i}] = 1 + p_i(\beta-1)$ where $p_i$ is the mean of the i'th bernoulli random variable. Now you observe

$$\prod_{i=1}^nE[\beta^{x_i}] = \prod_{i=1}^n(1 + p_i(\beta-1)) \leq \prod_{i=1}^n e^{p_i(\beta-1)}$$ $$= e^{(\beta - 1)\sum_{i=1}^n p_i} = e^{(\beta - 1)\mu}$$

Putting it all together, we get $$P(\sum_{i=1}^n x_i > \beta \mu) \leq e^{(-\beta \ln \beta +\beta -1)\mu} = e^{-g(\beta)\mu}$$


share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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