# Packages to determine confidence intervals

**URL:** <https://discourse.julialang.org/t/packages-to-determine-confidence-intervals/41233>\
**Category:** Statistics\
**Tags:** fit\
**Created:** [June 11, 2020, 10:37pm UTC](https://discourse.julialang.org/t/packages-to-determine-confidence-intervals/41233 "2020-06-11T22:37:55Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![feanor12](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/feanor12/32/8212_2.png) [@feanor12](https://discourse.julialang.org/u/feanor12)\
**Post date:** [June 11, 2020, 10:37pm UTC](https://discourse.julialang.org/t/packages-to-determine-confidence-intervals/41233/1 "2020-06-11T22:37:55Z")

</div>

As LsqFit.jl only calculates the standard error by using the Jacobian at the best-fit position I thought I try to write a small package for determining the confidence interval by data exploration.

[https://github.com/feanor12/ConfidenceInterval.jl](https://github.com/feanor12/ConfidenceInterval.jl)

As I am no expert I just implemented a method using an F-test and a simple linescan to test each parameter. (correlation is not taken into account yet, but creating a map in parameter space should not be that hard)

As this package is at a very early stage I wanted to check if something like this is already out there.

---

<div class="post-metadata">

**Author:** ![feanor12](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/feanor12/32/8212_2.png) [@feanor12](https://discourse.julialang.org/u/feanor12)\
**Post date:** [June 12, 2020, 11:16am UTC](https://discourse.julialang.org/t/packages-to-determine-confidence-intervals/41233/2 "2020-06-12T11:16:34Z")

</div>

I could find F-tests being applied for confidence interval estimation in an youtube video and by the python package lmfit, but I could not find a good paper for the procedure.

I am wondering if the F-test is even a valid approach. Maybe someone with a deeper knowledge in statistics can help me here.

Matlab for example provides confidence interval estimation based on the Wald test, log likelihood and bootstrapping. Maybe a Wald test would be a good option to implement as bootstrapping is already covered by Bootstrap.jl.