6
$\begingroup$

I am trying to solve a question which gives me a random variable with the distribution function below

$$ F(x) = 1 - \left(\frac{\mu}{x}\right)^{2n} $$

where $0 < \mu \le x < \infty$

I differentiate this to obtain the PDF

$$ f(x) = 2n \mu^{2n} x^{-2n-1} $$

At this point, I notice that the integral over the PDF does not sum to 1

$$ \int_{-\infty}^\infty f(x) dx = \int_{-\infty}^\infty 2n \mu^{2n} x^{-2n-1} dx = 2n\mu^{2n} \left[ \frac{x^{-2n}}{-2n} \right]_{-\infty}^\infty = \mu^{2n} \left[ x^{-2n} \right]_{-\infty}^\infty = 0 $$

Have I gone wrong somewhere above? This is part of a bigger question but the CDF is stated in the question as above.

$\endgroup$
7
  • 18
    $\begingroup$ 1. You are integrating from $-\infty$ to $\infty$, but the range of the variable is $[\mu,\infty)$. 2. You have dropped the $-$ sign in the last step of your integration. 3. Why are you taking the derivative and then integrating the result? You already have the functional form of $F(x)$! You can see by inspection that $F(\mu) = 0$ and $\lim_{x \to \infty}F(x) = 1$. $\endgroup$
    – jbowman
    Commented Jan 8, 2023 at 19:04
  • 2
    $\begingroup$ However, any estimator with zero variance is almost surely a constant, which means its distribution function cannot be this particular $F$! $\endgroup$
    – whuber
    Commented Jan 8, 2023 at 19:36
  • 1
    $\begingroup$ As far as the variance is concerned, the integral of $2x(1-F(x))\mathrm dx$ will compute it, assuming you correctly use the implied fact that $F(x)=0$ for all $x\le \mu.$ $\endgroup$
    – whuber
    Commented Jan 8, 2023 at 21:21
  • 1
    $\begingroup$ @Solomon Sometimes there are so many issues raised in a question that comments like jbowman's are effective ways to get to the point and understand how best to respond. $\endgroup$
    – whuber
    Commented Jan 9, 2023 at 14:31
  • 2
    $\begingroup$ @SolomonUcko - I thought about writing a full answer, but my third sub-comment was questioning whether the OP had gone down a pointless path in the approach to the entire problem rather than trying to point out mistakes, which I felt was better placed in a comment. Maybe I should have done both! $\endgroup$
    – jbowman
    Commented Jan 9, 2023 at 16:07

1 Answer 1

22
$\begingroup$

As pointed out in comments, the range of integration in your integral does not match the listed support of the random variable (which is $\mu \leqslant x < \infty$). Start by correcting the expression for your density, with explicit statement of the support:

$$f(x) = \begin{cases} \frac{2n}{\mu} \Big( \frac{\mu}{x} \Big)^{2n+1} & & & \text{for } x \geqslant \mu, \\[6pt] 0 & & & \text{otherwise}. \\[6pt] \end{cases}$$

Now correct your integral:

$$\begin{align} \int \limits_{-\infty}^\infty f(x) \ dx &= \int \limits_{-\infty}^\mu 0 \ dx + \int \limits_{\mu}^\infty \frac{2n}{\mu} \Big( \frac{\mu}{x} \Big)^{2n+1} \ dx \\[6pt] &= 2n \mu^{2n} \int \limits_{\mu}^\infty x^{-2n-1} \ dx \\[6pt] &= 2n \mu^{2n} \bigg[ -\frac{x^{-2n}}{2n} \bigg]_{x = \mu}^{x \rightarrow \infty} \\[6pt] &= 2n \mu^{2n} \bigg[ 0 - \Big( -\frac{\mu^{-2n}}{2n} \Big) \bigg] \\[6pt] &= 2n \mu^{2n} \cdot \frac{\mu^{-2n}}{2n} \\[12pt] &= 1. \\[6pt] \end{align}$$

$\endgroup$
1
  • $\begingroup$ I miss calculus. $\endgroup$
    – J. Doe.
    Commented Jan 10, 2023 at 14:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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