1
$\begingroup$

I would like to compute (and plot) the exact density of the following distribution:

$ X_i \sim exp(-Exponential(\lambda)) - 0.5 $

I already have the estimated density for this distribution, but I need to get its exact density to compute the MSE of my kernel estimator.

I have difficulties to transform the function to be able to plot its density.

For now, I have tried to compute it (with r). I used the inverse of the transformation to be able to compute the pdf using dexp().

X2 <- exp(-rexp(n,rate =1))-0.5 #get random sample
X2<-sort(X) 
plot(X2,dexp(-log(X2+0.5))) 

But this gives me results that are not coherent with my estimated density. comparison of true vs estimated

What am I doing wrong here?

$\endgroup$
  • 2
    $\begingroup$ Welcome to CV nahaut. To see what goes wrong, write $\text{Pr}[X \leq x ]$ as a probability involving the exponentilal r.v. say $E$ then use the known distribution of $E$. $\endgroup$ – Yves Jan 5 at 15:54
  • $\begingroup$ Thanks. I don't understand exactly what you meant. I tried to calculate the density using the transformation method but it's not giving me the correct results. $\endgroup$ – nahaut Jan 5 at 19:20
  • $\begingroup$ I considered that $ -ln(X_i + 0.5) \sim Exponential(\lambda)$. Then, I considered that it was a function of the random variable: $U= -ln(X_i + 0.5)$. After computation, the result I got was not giving the adequate density: $fu(U)= e^{-e^{-u}+0.5-u}$ $\endgroup$ – nahaut Jan 5 at 19:27
  • 2
    $\begingroup$ Note that $X$ has its values in $I:=(-0.5, \, 0.5)$. For $x \in I$, we have $\text{Pr}\{X \leq x \} = \text{Pr}\{U \geq - \log(x + 0.5) \} = \exp[\lambda \log(x + 0.5)]$ since $\text{Pr}\{U \geq u \} = \exp[-\lambda u]$ for $u > 0$. So $\text{Pr}\{X \leq x \} = [ x + 0.5]^\lambda$ and the density comes by derivation. Please use the appropriate tag homework̀ or self-study. $\endgroup$ – Yves Jan 6 at 8:15
  • 1
    $\begingroup$ Look for the keyword Jacobian to understand your mistake. $\endgroup$ – Xi'an Jan 6 at 21:25

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Browse other questions tagged or ask your own question.