0
$\begingroup$

I have a dataset and I checked that fits a Laplace distribution. I want to get different confidence intervals from it.

I know that in a normal distribution, the confidence interval of 68% is mean + std, but, what about Laplace distribution? Is it possible to convert it to a normal distribution? Does it make sense? How can I do it in python?

Improve this question
$\endgroup$
5
  • 1
    $\begingroup$ Please give us more details. What exactly you want to compute the confidence interval for? CI's are calculated for some estimated parameters, not for the whole datasets. You want to compute theoretical quantiles of the Laplace distribution given the parameters estimated from the dataset (then why not empirical quantiles)? $\endgroup$
    – Tim
    Commented Oct 29, 2018 at 10:14
  • $\begingroup$ And how can I calculate the empirical quantiles? (I am new in statistics, and I do not understand it well, sorry...). I know that in a normal distribution a confidence interval of 68% is equivalent to mean + std, but what about Laplace? Is it possible to calculate it? If it is not, can I convert it to a normal distribution, and then calculate mean + std? $\endgroup$
    – jartymcfly
    Commented Oct 29, 2018 at 10:18
  • $\begingroup$ If this is new and unclear for you, then could you please describe in plain English what you are trying to achieve? Why do you need to calculate the things you are trying to calculate? What do you need to learn from your data? $\endgroup$
    – Tim
    Commented Oct 29, 2018 at 10:21
  • $\begingroup$ I have a laplace distribution and I want to check which are the limit values for a confidence interval of 95%. I know that in normal distribution it is calculated: mean + 2std, but I don't really know how is it calculated in a laplace distribution. I don'w know if I have explained it well... $\endgroup$
    – jartymcfly
    Commented Oct 29, 2018 at 10:29
  • 1
    $\begingroup$ Then take the 0.025 and 0.975 quantiles of the Laplace distribution, i.e. $F^{-1}(p) = \mu - b\,\mathrm{sgn}(p-0.5)\,\ln(1 - 2|p-0.5|)$. But from your description it rather sounds like you have some data and want to learn something from the data, rather then estimating quantiles of theoretical distribution... $\endgroup$
    – Tim
    Commented Oct 29, 2018 at 10:32

0

Reset to default

Your Answer

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